CN114330444A - Circuit load identification method and device - Google Patents
Circuit load identification method and device Download PDFInfo
- Publication number
- CN114330444A CN114330444A CN202111633182.2A CN202111633182A CN114330444A CN 114330444 A CN114330444 A CN 114330444A CN 202111633182 A CN202111633182 A CN 202111633182A CN 114330444 A CN114330444 A CN 114330444A
- Authority
- CN
- China
- Prior art keywords
- data
- load
- neural network
- convolutional neural
- network model
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a circuit load identification method and a circuit load identification device, wherein the method comprises the following steps: acquiring current data and voltage data of a home circuit and capacitance characteristic information of a load, converting the current data and the voltage data into load image information of the home circuit based on the capacitance characteristic information, inputting the load image information into a preset target convolutional neural network model, obtaining a load scalar characteristic value of the home circuit, comparing the load scalar characteristic value with a characteristic value in a sample library, and obtaining the load category of the home circuit. The invention improves the accuracy of load identification through the circuit load identification method.
Description
Technical Field
The invention relates to the field of product detection, in particular to a circuit load identification method and device.
Background
With the continuous development of intelligent technology and the increasing emphasis of the public on energy-saving methods, the public has more emphasis on the fields of energy management and power monitoring in recent years, and as an important part of the fields, the optimization of household power utilization modes has great significance on the progress of the fields, and is regarded as a way capable of improving energy consumption. Among them, identification of household loads is an indispensable technique for improving household power consumption. The on-line identification of the household appliances is an unavoidable task for optimizing the household electricity utilization mode. The intrusive load identification method for installing the sensing equipment on a single load needs expensive hardware support, the non-intrusive load identification only needs to install the sensing sampling identification equipment at the main entrance of household power transmission, and the non-intrusive household load identification has more economic significance and practical value.
Common non-intrusive home load identification algorithms fall into the event detection class-based and the non-event detection class-based. The recognition algorithm not based on the event detection class recognizes the combination of a plurality of loads in the online state, and the recognition class increases at double speed with the increase of the load type, so that the practical significance is not great. The load of the switching event is mainly identified based on the event detection class, and single-load sample data can be obtained by carrying out difference on the same reference.
There are still a number of recognized difficulties in identifying in the case of single load samples: under the background of multiple types and brands of household appliances, the identification accuracy rate is difficult to improve for the traditional machine learning algorithm, so that the methods are not mature in practice; in recent years, although the recognition accuracy is high, a sufficient amount of samples and time training are often required for recognition by a hot deep learning network model method. In addition, the recognition algorithms have insufficient learning capacity at one time, and are difficult to meet the actual use condition of the household electrical appliances.
Therefore, in order to improve the accuracy of load identification and solve the technical problem that the accuracy of load identification by the existing traditional learning algorithm is low, it is urgently needed to construct an identification method of circuit load.
Disclosure of Invention
The invention provides a circuit load identification method and a circuit load identification device, which solve the technical problem that the accuracy of load identification is low in the existing traditional learning algorithm.
In a first aspect, the present invention provides a method for identifying a circuit load, including:
acquiring current data and voltage data of a household circuit and capacitance and inductance characteristic information of a load;
converting the current data and the voltage data into load image information of the household circuit based on the capacitance characteristic information;
inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the household circuit;
and comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the home circuit.
Optionally, converting the current data and the voltage data into image information of the load based on the capacitive sensing characteristic information includes:
converting the current data and the voltage data into track image data;
and adding the appearance characteristic information into the track image data to obtain the image information of the load.
Optionally, inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the home circuit, where the method includes:
dividing the load image information into training data and verification data;
training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model;
and inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain a load scalar characteristic value of the household circuit.
Optionally, training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model, including:
inputting the training data into the preliminary convolutional neural network model, and training to obtain a trained convolutional neural network model;
and verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
Optionally, inputting the training data into the preliminary convolutional neural network model, and performing training to obtain a trained convolutional neural network model, including:
setting parameter data of the preliminary convolutional neural network model;
inputting the training data into the preliminary convolutional neural network model to obtain training result data;
and optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model.
In a second aspect, the present invention provides an apparatus for identifying a circuit load, comprising:
the acquisition module is used for acquiring current data and voltage data of the household circuit and capacitance and inductance characteristic information of a load;
the image module is used for converting the current data and the voltage data into load image information of the household circuit based on the capacitance characteristic information;
the input module is used for inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the family circuit;
and the comparison module is used for comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the household circuit.
Optionally, the image module comprises:
the image submodule is used for converting the current data and the voltage data into track image data;
and the characteristic submodule is used for adding the capacitance characteristic information into the track image data to obtain the image information of the load.
Optionally, the input module comprises:
the dividing submodule is used for dividing the load image information into training data and verification data;
the training submodule is used for training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model;
and the input submodule is used for inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain the load scalar characteristic value of the household circuit.
Optionally, the training submodule includes:
the training unit is used for inputting the training data into the preliminary convolutional neural network model for training to obtain a trained convolutional neural network model;
and the verification unit is used for verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
Optionally, the training unit comprises:
the parameter subunit is used for setting parameter data of the preliminary convolutional neural network model;
the input subunit is used for inputting the training data into the preliminary convolutional neural network model to obtain training result data;
and the optimizing subunit is used for optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model.
According to the technical scheme, the invention has the following advantages: the invention provides a circuit load identification method, which comprises the steps of converting current data and voltage data of a household circuit into load image information of the household circuit by acquiring current data and voltage data of the household circuit and capacitance characteristic information of a load based on the capacitance characteristic information, inputting the load image information into a preset target convolution neural network model to obtain a load scalar characteristic value of the household circuit, comparing the load scalar characteristic value with a characteristic value in a sample library to obtain a load category of the household circuit.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art 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 can be obtained according to these drawings without inventive exercise.
FIG. 1 is a flowchart illustrating a first embodiment of a method for identifying a circuit load according to the present invention;
FIG. 2 is a flowchart illustrating a second embodiment of a circuit load identification method according to the present invention;
fig. 3 is a block diagram of an intelligent home system according to the present invention;
FIG. 4 is a diagram illustrating image information of a load in a database according to the present invention;
FIG. 5 is a schematic diagram showing the comparison of the capacitance information of two loads in the image according to the present invention;
FIG. 6 is a schematic view of an identification training sub-interface of an intelligent home system according to the present invention;
FIG. 7 is a schematic view of a voltage-current display sub-interface of an intelligent home system of the present invention;
FIG. 8 is a schematic diagram of a trajectory imaging sub-interface of an intelligent home system according to the present invention;
FIG. 9 is a graph of training accuracy of a convolutional neural network model of the present invention as a function of iteration number;
FIG. 10 is a graph of the validation accuracy of a convolutional neural network model of the present invention as a function of iteration number;
fig. 11 is a block diagram of an embodiment of a circuit load recognition apparatus according to the present invention.
Detailed Description
The embodiment of the invention provides a circuit load identification method and device, which are used for solving the technical problem that the load identification accuracy of the traditional learning algorithm is low at present.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a first embodiment of a method for identifying a circuit load according to the present invention, including:
step S101, acquiring current data and voltage data of a household circuit and capacitance and inductance characteristic information of a load;
step S102, converting the current data and the voltage data into load image information of the household circuit based on the capacitance characteristic information;
step S103, inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the family circuit;
and step S104, comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the household circuit.
According to the circuit load identification method provided by the embodiment of the invention, the current data and the voltage data of the household circuit and the capacitance characteristic information of the load are obtained, based on the capacitance characteristic information, the current data and the voltage data are converted into the load image information of the household circuit, the load image information is input into a preset target convolutional neural network model, the load scalar characteristic value of the household circuit is obtained and compared with the load scalar characteristic value and the characteristic value in a sample library, and the load category of the household circuit is obtained.
In a second embodiment, referring to fig. 2, fig. 2 is a flowchart illustrating a method for identifying a circuit load according to the present invention, including:
step S201, acquiring current data and voltage data of a household circuit and capacitance and inductance characteristic information of a load;
in the embodiment of the invention, the current data and the voltage data of the household circuit and the capacitance and inductance characteristic information of the load are obtained.
In a specific implementation, please refer to fig. 3, where fig. 3 is a block diagram of a smart home system according to the present invention; the method comprises the following steps of firstly, obtaining a database 301, obtaining a data layer 302, obtaining an algorithm layer 303, obtaining a hidden layer 304, obtaining a GUI (graphical user interface) 305, obtaining a GUI 306, obtaining an original database PLAID, obtaining an event detection 307, obtaining a V-I track conversion 308, obtaining a track image 309, obtaining features 310, obtaining load identification 311, obtaining an identification result processing 312, obtaining an identification process 313, obtaining a training process 314, obtaining a V-I track and related display 315, obtaining data to be identified by pouring 316, obtaining new electrical appliance sample updating 317, obtaining identification display 318, and obtaining a sample library 319.
Three levels and a constantly updated sample repository 319 are in the middle of the functional implementation of the GUI interactive interface 305 from the original database plidi 306 to the last where the load can be identified in real time. The data layer 302 mainly implements processing of original data, and the main function is to implement imaging of data. The algorithm layer 303 performs the functions of feature extraction 310 and load identification 311 on the algorithm level for the generated image, and is a core part of the whole design process. There is an interactive process between the GUI interactive interface 305 portion and the algorithm layer 303, namely a training process and a recognition process. It operates manually through the GUI interface 305 on-line, and the off-line process is performed through the algorithm layer 303. In this process, the recognition training process requires the input of the sample library to be given, and the discovery of a new sample updates the sample library. The real-time updated sample library can effectively improve the identification accuracy, and the training process of the user side becomes possible.
In practice, common data for non-intrusive home load identification is steady-state, transient current-voltage data, which is typically the concurrent operation of multiple like or heterogeneous loads. According to the characteristics of parallel connection of household circuits, the relationship between the total current and each branch current is as follows:
wherein, IGeneral assemblyA variable representing a varying total current; i isiA variable representing a single load current that varies continuously; n represents the total number of online loads.
The relationship between the total voltage and the branch line voltages is:
Ugeneral assembly＝U1＝...＝UN
Due to the above relation between the total current voltage and the branch line current voltages, when the online combination of the loads under the total steady state condition is difficult to identify, the single load data of the input and the cut-off can be extracted to identify the loads under the condition that the input and the cut-off of the loads are detected by the event detection program.
The data extraction method comprises the following steps: after the input cutting event is detected, taking current data of ten periods backwards in a two-end steady state process before and after the input cutting event by taking a voltage highest point as a reference initial point (ensuring that the ten periods are in the steady state process), and then carrying out difference according to the input cutting event to obtain data of ten periods of single load.
Step S202, converting the current data and the voltage data into track image data;
in an embodiment of the present invention, the current data and the voltage data are converted into trajectory image data.
In the specific implementation, firstly, the current and voltage values of the same sample of the PLAID data set are synchronously acquired, and the voltage information has periodicity, so that the voltage is used as a reference, the period from the maximum amplitude to the minimum amplitude of the voltage is the first half period, the period from the minimum amplitude to the maximum amplitude is the second half period, and the position of a black or blue pixel point is determined by the corresponding current value.
In order to realize imaging, the current and voltage are normalized, and the obtained current and voltage range is (-1, 1). The size of the image to be obtained can be designed, and an image size of 64 × 3 is selected. The size of each pixel is determined by the value range of the voltage and the current. The characteristic is embodied in that each row and each column of the image must have a pixel point which is not (255 ).
Step S203, adding the appearance characteristic information into the track image data to obtain the image information of the load;
in the embodiment of the invention, the appearance characteristic information is added to the track image data to obtain the image information of the load.
In the implementation, please refer to fig. 4, fig. 4 is a schematic diagram of image information of a load in a database according to the present invention; in order to embody the capacitance characteristic of the load, black pixel points, namely (0,0,0), are taken from track points of the previous half period; and taking blue pixel points, namely (0, 255), from the trace points of the second half period. The rules for expressing the appearance are as follows: from the highest voltage point to the lowest voltage point and then to the highest voltage point, the trace direction of firstly black and then blue is clockwise, and the capacitive load is formed; counterclockwise is an inductive load.
In order to avoid the influence of random interference on the identification result, the current and voltage data of ten steady periods are adopted for data conversion into images. In actual operation, ten steady-state periods are respectively imaged, and the final images are superposed to obtain a steady-state V-I track image, wherein a black pixel point is represented by a number 1 in a gray image, and a blue pixel point is represented by a number 2 to indicate distinction.
In order to further improve the information of the load, the information of the appearance and the feeling of the load is added into the V-I image, and the information is embodied into the motion direction of the V-I track when the color moves.
Referring to fig. 5, fig. 5 is a schematic diagram illustrating a comparison of capacitance-sensitivity information of two loads in an image, wherein a fan image represented by a left-side diagram shows an inductive characteristic, a washing machine image represented by a right-side diagram shows a capacitive characteristic, in order to distinguish blue and black pixels, the numbers 1 and 2 are marked in the diagram, an image of a rectangular half area where 1 is located represents a black pixel, and an image of a rectangular half area where 2 is located represents a blue pixel.
Step S204, dividing the load image information into training data and verification data;
step S205, training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain a target convolutional neural network model;
it should be noted that the parameter data is a variable of the convolutional neural network model, and the parameter is adjusted as needed, so that the convolutional neural network model can be optimized to obtain the target convolutional neural network model.
The data labels corresponding to the training data are the results of the training data corresponding to the reality, the training results after training are compared with the reality results, parameters are optimized, and the accuracy of the convolutional neural network model can be improved.
In an optional embodiment, training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model includes:
setting parameter data of the preliminary convolutional neural network model;
inputting the training data into the preliminary convolutional neural network model to obtain training result data;
optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model;
and verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
In the embodiment of the invention, parameter data of a preliminary convolutional neural network model is set firstly, then training data is input into the preliminary convolutional neural network model to obtain training result data, parameters are adjusted according to errors of data labels corresponding to the training data and the training result data to obtain optimal parameters, and the convolutional neural network model is optimized by utilizing the optimal parameters to obtain the trained neural network model.
In a specific implementation, the Siamese convolutional neural network is a network structure for comparing image similarity, and has two branches. The network parameters of two branches are shared, and a single branch is a multilayer convolutional neural network. And inputting the generated V-I image into a classification sub-network through image features extracted by a convolutional neural network, and training the network under the supervision of a classification loss function.
The classification subnetwork treats the follow-up work as a binary task with the purpose of maximizing inter-class differences and minimizing intra-class differences. The loss function used here is a cross-entropy loss function. The training algorithm of the convolutional neural network of the feature extraction part is a back propagation random gradient descent method.
Compared with the training of a single convolutional neural network on the whole samples, the Simese convolutional neural network is more similar to the one-by-one training of the single samples rather than the overall training of the whole samples, and on the basis, the one-time learning capability of the Simese convolutional neural network is improved.
Step S206, inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain a load scalar characteristic value of the household circuit;
step S207, comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the household circuit;
in the embodiment of the invention, the load scalar characteristic value is compared with the characteristic value in the sample library to obtain the load category of the home circuit.
According to the circuit load identification method provided by the embodiment of the invention, the current data and the voltage data of the household circuit and the capacitance characteristic information of the load are obtained, based on the capacitance characteristic information, the current data and the voltage data are converted into the load image information of the household circuit, the load image information is input into a preset target convolutional neural network model, the load scalar characteristic value of the household circuit is obtained and compared with the load scalar characteristic value and the characteristic value in a sample library, and the load category of the household circuit is obtained.
Referring to fig. 6, fig. 6 is a schematic view of an identification training sub-interface of an intelligent home system according to the present invention, fig. 7 is a schematic view of a voltage and current display sub-interface of an intelligent home system according to the present invention, fig. 8 is a schematic view of a trajectory imaging sub-interface of an intelligent home system according to the present invention, and a description of nesting of a GUI interactive interface and an algorithm is given by combining fig. 6, fig. 7, and fig. 8.
Selecting 8 types of electric appliances (Air Conditioner, Compact Fluorescent Lamp, Fan, Fridge, Hairdyer, Heat, Incandesce Light Bulb, Laptop) which do not participate in training in a sample library to identify:
"select load file to be identified" in the interface shown in fig. 6: selecting a sample point voltage and current file to be identified in a drop-down list, clicking a key for starting identification if the Air Conditioner _1.csv is selected, reading the file by a GUI background, extracting voltage and current data, and displaying voltage and current waveforms on a voltage/current waveform interface. Meanwhile, the background inputs the voltage and current data into an event detection program, and the principle of the program is that current data of 20 periods are read in sequence each time, if the change of the effective value of the current of the next 10 periods relative to the effective value of the current of the previous 10 periods is greater than 0.01A, a resampling program is triggered, the position of the last sampling point of the 20 periods is found, the positive value after the first zero crossing point of the voltage is found from the next point of the point, the voltage and current sequence of ten periods is taken from the positive value, namely the steady-state voltage and current data of the ten periods, and the ten steady-state voltage and current waveforms are displayed on a voltage/current waveform interface, as shown in fig. 7. Then, the background displays the voltage-current traces formed by the ten steady-state voltage and current data on the "voltage-current trace" interface, as shown in fig. 7, and simultaneously, the background converts the voltage-current traces formed by the ten steady-state voltage and current data into RGB images and displays the RGB images on the "voltage-current trace" interface, as shown in fig. 8. Finally, the background inputs the generated RGB images and the images in the sample library (stored in the "sample _ configuration" folder) into the recognition program, and the recognition program matches the RGB images and the images in the sample library one by one and inputs the RGB images and the images in the sample library into the trained model to obtain a recognition result and the possibility of the result, and outputs and displays the result on the interface of "loading the electrical appliance to be recognized", as shown in fig. 6.
Newly adding an electric appliance to the sample library and realizing the identification of the newly added electric appliance:
when an electrical appliance needs to be added to a sample library, selecting a 'Browse to File.' key (or directly filling an address in a rear blank) in an interface for loading the electrical appliance to be identified, selecting a voltage and current sampling data File (. csv) of the newly added electrical appliance, selecting a 'Copy File To..' key (or directly filling an address in the rear blank), selecting a sample library address, storing a default sample library in a 'sample _ library' folder, clicking a 'Copy' key after selection, adding the voltage and current sampling data File (. csv) of the newly added electrical appliance into the sample library by a background, and then adding the name of the newly added electrical appliance into the sample library: "fill the name of the new electrical apparatus in the blank place behind, click" the new electrical apparatus name is: the background will name the newly added file in the sample library as a uniform format (appliance name + "_" + str (the file serial number) + ". csv)" by the key, and at the same time, the background will generate the RGB image from the file and store it in the "sample _ configuration" folder. At this point, the new appliance has been added to the sample library.
Referring to fig. 9, fig. 9 is a graph illustrating a training accuracy of a convolutional neural network model according to the present invention varying with iteration times, and fig. 10 is a graph illustrating a verification accuracy of a convolutional neural network model according to the present invention varying with iteration times;
8 electric appliances (Air Conditioner, Compact Fluorescent Lamp, Fan, Fridge, Hairdyer, Heater, Incandecent Light Bulb, Laptop) in the 11 electric appliances in the PLAID database are selected to train and test the network, wherein six groups of data are selected for each electric appliance to serve as a training set, and four groups of data serve as a cross validation set. And selecting other three electric appliances (Microwave, Vacuum, Washing Machine) to test the learning ability of the trained model for one time. The curve of the training accuracy rate along with the iteration times of the model in the training process is shown in fig. 9, the curve of the verification accuracy rate along with the iteration times is shown in fig. 10, wherein the accuracy rate of the model for identifying the 8 electric appliances is up to 98% through testing on a cross verification set, the accuracy rate of the model for identifying the 8 electric appliances is up to 97.58% through testing on a test set, and the identification accuracy rate for each electric appliance is shown in table 1.
TABLE 1
The analysis of table 1 shows that although the air conditioner, the fan, the water heater and the incandescent lamp have similar shapes, the recognition accuracy of the air conditioner, the fan, the water heater and the incandescent lamp is as high as 100%, because the electric appliances with the voltage-current tracks occupy half (4 types) of the training set, the recognition accuracy of the model for the electric appliances is greatly improved, and because the waveforms of the compact fluorescent lamp and the notebook computer are similar and the proportion of the electric appliances in the training set is not large (only 2 types, accounting for 25%), the recognition accuracy of the compact fluorescent lamp and the notebook computer is reduced compared with that of other electric appliances, but the recognition accuracy is still higher.
In the embodiment of the invention, the three electric appliances (a microwave oven, a dust collector and a washing machine) which do not participate in training are subjected to one-time learning ability test on a trained model, and the result shows that the identification accuracy of the microwave oven is 95.56%, the identification accuracy of the dust collector is 94.44%, the identification accuracy of the washing machine is 91.11%, and the overall identification accuracy is as high as 95.32%. The result shows that the model has better one-time learning ability.
The method comprises the following steps of carrying out learning capability test on three electric appliances (a microwave oven, a dust collector and a washing machine) which do not participate in training on a GUI interface designed by people, sequentially adding voltage and current sampling data of the three electric appliances of the microwave oven, the dust collector and the washing machine into a sample library, respectively inputting corresponding electric appliance names, and clicking' the new electric appliance name is as follows: and the GUI draws a corresponding voltage-current track image in the background and stores the image in a sample _ configuration folder, so that the newly added electric appliance can be identified. The result shows that 10 groups of data of 12 groups of data in total of the three electric appliances are accurately identified, and the learning capability at one time is better.
Referring to fig. 11, fig. 11 is a block diagram illustrating an embodiment of a circuit load identification apparatus according to the present invention, including:
an obtaining module 1101, configured to obtain current data and voltage data of a home circuit, and capacitance-inductance characteristic information of a load;
an image module 1102, configured to convert the current data and the voltage data into load image information of the home circuit based on the capacitive sensing characteristic information;
an input module 1103, configured to input the load image information to a preset target convolutional neural network model, so as to obtain a load scalar characteristic value of the home circuit;
and the comparison module 1104 is used for comparing the load scalar characteristic value with the characteristic values in the sample library to obtain the load category of the home circuit.
In an alternative embodiment, the image module 1102 includes:
the image submodule is used for converting the current data and the voltage data into track image data;
and the characteristic submodule is used for adding the capacitance characteristic information into the track image data to obtain the image information of the load.
In an alternative embodiment, the input module 1103 includes:
the dividing submodule is used for dividing the load image information into training data and verification data;
the training submodule is used for training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model;
and the input submodule is used for inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain the load scalar characteristic value of the household circuit.
In an alternative embodiment, the training submodule includes:
the training unit is used for inputting the training data into the preliminary convolutional neural network model for training to obtain a trained convolutional neural network model;
and the verification unit is used for verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
In an alternative embodiment, the training unit comprises:
the parameter subunit is used for setting parameter data of the preliminary convolutional neural network model;
the input subunit is used for inputting the training data into the preliminary convolutional neural network model to obtain training result data;
and the optimizing subunit is used for optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present application, it should be understood that the method and apparatus disclosed in the present invention can be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, 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.
The 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 unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a readable storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned readable storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will 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; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for identifying a circuit load, comprising:
acquiring current data and voltage data of a household circuit and capacitance and inductance characteristic information of a load;
converting the current data and the voltage data into load image information of the household circuit based on the capacitance characteristic information;
inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the household circuit;
and comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the home circuit.
2. The method for identifying a circuit load according to claim 1, wherein converting the current data and the voltage data into image information of the load based on the capacitive sensing characteristic information comprises:
converting the current data and the voltage data into track image data;
and adding the appearance characteristic information into the track image data to obtain the image information of the load.
3. The method for identifying circuit loads according to claim 1, wherein inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the home circuit comprises:
dividing the load image information into training data and verification data;
training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model;
and inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain a load scalar characteristic value of the household circuit.
4. The method for identifying circuit loads according to claim 3, wherein training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model comprises:
inputting the training data into the preliminary convolutional neural network model, and training to obtain a trained convolutional neural network model;
and verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
5. The method for identifying circuit loads according to claim 4, wherein the step of inputting the training data into the preliminary convolutional neural network model for training to obtain a trained convolutional neural network model comprises:
setting parameter data of the preliminary convolutional neural network model;
inputting the training data into the preliminary convolutional neural network model to obtain training result data;
and optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model.
6. An apparatus for identifying a circuit load, comprising:
the acquisition module is used for acquiring current data and voltage data of the household circuit and capacitance and inductance characteristic information of a load;
the image module is used for converting the current data and the voltage data into load image information of the household circuit based on the capacitance characteristic information;
the input module is used for inputting the load image information into a preset target convolutional neural network model to obtain a load scalar characteristic value of the family circuit;
and the comparison module is used for comparing the load scalar characteristic value with the characteristic value in the sample library to obtain the load category of the household circuit.
7. The apparatus for circuit load recognition according to claim 6, wherein the image module comprises:
the image submodule is used for converting the current data and the voltage data into track image data;
and the characteristic submodule is used for adding the capacitance characteristic information into the track image data to obtain the image information of the load.
8. The apparatus for circuit load recognition according to claim 6, wherein the input module comprises:
the dividing submodule is used for dividing the load image information into training data and verification data;
the training submodule is used for training and verifying a preliminary convolutional neural network model based on the training data and the verification data to obtain the target convolutional neural network model;
and the input submodule is used for inputting the load image information of the household circuit to be tested into the target convolutional neural network model to obtain the load scalar characteristic value of the household circuit.
9. The apparatus for circuit load recognition according to claim 8, wherein the training submodule comprises:
the training unit is used for inputting the training data into the preliminary convolutional neural network model for training to obtain a trained convolutional neural network model;
and the verification unit is used for verifying the trained convolutional neural network model based on the verification data to obtain the target convolutional neural network model.
10. The apparatus for circuit load recognition according to claim 9, wherein the training unit comprises:
the parameter subunit is used for setting parameter data of the preliminary convolutional neural network model;
the input subunit is used for inputting the training data into the preliminary convolutional neural network model to obtain training result data;
and the optimizing subunit is used for optimizing the preliminary convolutional neural network model based on the parameter data, the data labels corresponding to the training data and the training result data to obtain the trained convolutional neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111633182.2A CN114330444A (en) | 2021-12-28 | 2021-12-28 | Circuit load identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111633182.2A CN114330444A (en) | 2021-12-28 | 2021-12-28 | Circuit load identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114330444A true CN114330444A (en) | 2022-04-12 |
Family
ID=81015692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111633182.2A Pending CN114330444A (en) | 2021-12-28 | 2021-12-28 | Circuit load identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114330444A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115952447A (en) * | 2023-03-14 | 2023-04-11 | 广东云山能源科技有限公司 | Intelligent identification system and method for load types of electric appliances |
-
2021
- 2021-12-28 CN CN202111633182.2A patent/CN114330444A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115952447A (en) * | 2023-03-14 | 2023-04-11 | 广东云山能源科技有限公司 | Intelligent identification system and method for load types of electric appliances |
| CN115952447B (en) * | 2023-03-14 | 2023-06-02 | 广东云山能源科技有限公司 | Intelligent identification system and method for electric appliance load type |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Non-intrusive load monitoring by voltage–current trajectory enabled transfer learning | |
| Basu et al. | Time series distance-based methods for non-intrusive load monitoring in residential buildings | |
| US8756181B2 (en) | System and method employing a self-organizing map load feature database to identify electric load types of different electric loads | |
| CN110443302A (en) | Load discrimination method and its application based on Fusion Features and deep learning | |
| US20130138669A1 (en) | System and method employing a hierarchical load feature database to identify electric load types of different electric loads | |
| CN110781332A (en) | Electric power resident user daily load curve clustering method based on composite clustering algorithm | |
| Abello et al. | A modular degree-of-interest specification for the visual analysis of large dynamic networks | |
| US20110137876A1 (en) | Data quality enhancement for smart grid applications | |
| CN111177216B (en) | Association rule generation method and device for comprehensive energy consumer behavior characteristics | |
| CN109885728B (en) | Video abstraction method based on meta-learning | |
| Niu et al. | Fitting‐based optimisation for image visual salient object detection | |
| CN107545049A (en) | Image processing method and related product | |
| CN114330444A (en) | Circuit load identification method and device | |
| Cui et al. | Feature extraction and classification method for switchgear faults based on sample entropy and cloud model | |
| CN115618249A (en) | Low-voltage power distribution station area phase identification method based on LargeVis dimension reduction and DBSCAN clustering | |
| Yin et al. | Research on non-intrusive unknown load identification technology based on deep learning | |
| CN113345052B (en) | Classification data multi-view visualization coloring method and system based on similarity significance | |
| CN115100466A (en) | Non-invasive load monitoring method, device and medium | |
| CN116881836A (en) | Non-invasive load identification method and system based on Gram angle field and feature fusion | |
| CN109918186A (en) | A kind of switching method of application program, device and intelligent terminal | |
| Jin et al. | Power load curve clustering algorithm using fast dynamic time warping and affinity propagation | |
| de Paula Rodrigues et al. | Curvature scale space-based signatures for electrical load classification in NILM | |
| Zhu et al. | Text segmentation using superpixel clustering | |
| CN110032585A (en) | A kind of time series bilayer symbolism method and device | |
| CN116108377A (en) | Non-invasive load state identification method and system based on hierarchical classification learning |
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 |