CN214473641U - Non-invasive load recognition device based on track image recognition - Google Patents

Abstract

The utility model belongs to the technical field of with electrical load discernment, a non-invasive load recognition device based on orbit image recognition is disclosed, non-invasive load recognition device based on orbit image recognition includes: the device comprises a current transformer, a voltage transformer, an electric energy metering chip, a main control chip, an AI processor, wireless communication equipment and a cloud server. The utility model adopts the electromagnetic sensor, which avoids the interference to the original electric system; the installation is simple, and the original circuit of the power utilization system is not required to be changed; the electrical characteristics of the electrical appliance are converted into a V-I track graph, and the type of the electrical appliance is more accurately identified by utilizing a mature image identification technology; meanwhile, compared with the existing electrical appliance feature library, the V-I track graph library is simpler to establish and maintain.

Description

Non-invasive load recognition device based on track image recognition

Technical Field

The utility model belongs to the technical field of with electrical load recognition, especially, relate to a non-invasive load recognition device based on orbit image recognition.

Background

At present, the application range of non-invasive electrical load recognition technology is wider, can be applied to intelligent apartment, wisdom house, can know the constitution condition of resident's power consumption through data analysis, accomplishes demand side response etc. and the user can know the power consumption custom of oneself simultaneously, improves the energy-conserving consciousness of oneself. However, in the prior art, a low-frequency load acquisition mode is mostly adopted, so that the identification precision is low, the real-time performance is poor, and the load change cannot be reflected in time; the high-frequency acquisition mode can cause large data transmission quantity and strong network dependence; meanwhile, the traditional load feature library is complex to establish, load features need to be extracted manually, and if the previous-stage feature extraction is not complete, the maintenance workload of the subsequent feature library is increased. Therefore, a new non-intrusive load identification system is needed.

Through the above analysis, the problems and defects of the prior art are as follows: in the prior art, a low-frequency load acquisition mode is adopted, so that the identification precision is low, the real-time performance is poor, and the load change cannot be reflected in time; the high-frequency acquisition mode can cause large data transmission quantity and strong network dependence; meanwhile, the traditional load feature library is complex to establish, load features need to be extracted manually, and if the previous-stage feature extraction is not complete, the maintenance workload of the subsequent feature library is increased.

The difficulty and significance for solving the problems and defects are as follows: by combining the AIOT technology, the load data acquisition and the load identification are both placed at the edge of the data acquisition, so that the uploading of a large amount of data is avoided; the traditional feature library is converted into a V-I track image, and one V-I track image can contain more load feature information, so that the maintenance work of the load feature library is reduced; meanwhile, by combining cloud service, a user can continuously perfect a load V-I track image library, so that more types of electrical appliances can be identified.

SUMMERY OF THE UTILITY MODEL

To the problem that prior art exists, the utility model provides a non-invasive load recognition device based on orbit image recognition.

The utility model discloses a realize like this, a non-invasive load recognition device based on orbit image recognition, include: the device comprises a current transformer, a voltage transformer, an electric energy metering chip, a main control chip, an AI processor, wireless communication equipment and a cloud server.

The household power consumption monitoring system comprises a current transformer, a voltage transformer and an electric energy metering chip, wherein the current transformer, the voltage transformer and the electric energy metering chip are arranged at an inlet of household power consumption, a sensor is arranged at the inlet of the household power consumption and is used for collecting current and voltage values of the household power consumption, the electric energy metering chip converts analog quantities output by the current and voltage sensors into digital quantities, counts active power and reactive power, and transmits real-time current, voltage values and power values to a main control chip;

the main control chip adopts a low-power consumption MCU as the core of the data acquisition system, processes data generated by the current transformer, the voltage transformer and the electric energy metering chip, judges whether an electric appliance switching event occurs according to the power change condition, generates a V-I track image of the switching electric appliance according to the current and voltage data if the electric appliance switching occurs, and transmits the generated V-I track image to the AI processor;

AI processor by I²C is connected with a main control chip, a special AI processor is adopted, high-speed convolutional neural network calculation can be carried out under ultra-low power consumption, and a V-I locus diagram can be rapidly identified to obtain a result;

the wireless communication equipment is used for transmitting the identification result output by the AI processor to the cloud server database in a network mode;

the cloud server mainly has a database and an AI learning function, wherein the database stores the electricity consumption data uploaded by the edge end and is used for data access of an application end, and the AI learning function is used for training a V-I track diagram of the electric appliance to generate a neural network model and downloading the neural network model to an AI processor through wireless communication equipment.

Further, the non-invasive load identification device based on track image identification further comprises:

the current sampling circuit monitors electricity consumption data of current, voltage and power in real time and the electric energy acquisition chip monitors the electricity consumption data of the current, the voltage and the power in real time;

the electric energy acquisition chip is provided with an SPI communication interface and transmits data to the main control chip through the SPI interface.

The non-invasive load identification device based on track image identification further comprises:

and the smart phone is connected with the cloud server through wireless communication equipment, such as WEB and a mobile phone APP.

Combine foretell all technical scheme, the utility model discloses the advantage that possesses and positive effect are: the utility model provides a non-invasive load recognition device based on V-I orbit image recognition combines thing networking and edge calculation to non-invasive mode monitoring consumer's in the certain limit in service behavior, near the sensor end through the V-I orbit picture that draws switching electrical apparatus, with the kind of the current switching electrical apparatus of mode recognition, need not to send a large amount of data to cloud ware analysis, reduce data transmission's requirement, and can look over the power consumption and the real-time status of each consumer at cell-phone PC end remote control. The utility model aims at monitoring the electricity consumption condition in real time, recording the electricity consumption data of each electric appliance and identifying the current online electric appliance type in real time; meanwhile, the identification algorithm is completed at the equipment end, and a large amount of current and voltage data do not need to be uploaded to a cloud server for analysis, so that the dependence on the network is reduced, and the real-time performance of identification is improved. The beneficial effects of the utility model also include:

(1) the utility model discloses an electromagnetic type sensor need not to insert and treats the monitoring system in, has avoided producing the interference to former electric system to the installation is simple, need not change former electric system circuit.

(2) The utility model discloses can be through internet of things real-time detection power consumption electrical apparatus state, can long-rangely look over opening of electrical apparatus in the current family and stop the state, avoid appearing the unexpected condition of opening or forgetting the closing of certain electrical apparatus, the user can know the in service behavior of electrical apparatus in the family simultaneously, improves the energy-conserving consciousness of oneself with data, reduces domestic power consumption to the utmost.

(3) The utility model discloses with load recognition algorithm operation near the sensor end, need not to send a large amount of data to the analysis of cloud ware, reduce data transmission's requirement.

(4) The utility model discloses a turn into the electrical characteristics of electrical apparatus V-I track map, utilize ripe image recognition technology, discern the electrical apparatus kind more accurately, V-I track map storehouse is compared with current electrical apparatus characteristic storehouse simultaneously, and it is simpler with the maintenance work to establish.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained from the drawings without creative efforts.

Fig. 1 is a schematic overall view of a non-invasive load recognition apparatus based on V-I trajectory image recognition according to an embodiment of the present invention.

Fig. 2 is a block diagram of a non-intrusive load recognition device based on V-I trajectory image recognition according to an embodiment of the present invention.

Fig. 3 is a structural diagram of an identification device according to an embodiment of the present invention.

Fig. 4 is a diagram of a neural network model structure for V-I trace map recognition according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

To solve the problems in the prior art, the present invention provides a non-invasive load recognition system and method based on V-I track image recognition, and the following description is made in detail with reference to the accompanying drawings.

The utility model discloses at first, provide a non-invasive load identification system based on V-I orbit image recognition, include:

the household electricity consumption monitoring system comprises a current transformer, a voltage transformer and an electric energy metering chip, wherein the current transformer, the voltage transformer and the electric energy metering chip are mainly composed of the current transformer, the voltage transformer and the electric energy metering chip, a sensor is installed at an inlet (such as the vicinity of an ammeter) of household electricity consumption and is used for collecting current and voltage values of the household electricity consumption as shown in figure 1, the electric energy metering chip converts analog quantities output by the current and voltage sensors into digital quantities, counts active power and reactive power, and transmits real-time current, voltage values and power values to a main control chip;

the main control chip adopts a low-power consumption MCU as the core of the data acquisition system, processes data generated by the current transformer, the voltage transformer and the electric energy metering chip, judges whether an electric appliance switching event occurs according to the power change condition, generates a V-I track image of the switching electric appliance according to the current and voltage data if the electric appliance switching occurs, and transmits the generated V-I track image to the AI processor;

the AI processor adopts a special AI processor, can perform high-speed convolutional neural network calculation under ultra-low power consumption, can rapidly identify the V-I locus diagram to obtain a result, and can obtain the result through I²C is connected with the main control chip;

and the wireless communication equipment is responsible for transmitting the identification result output by the AI processor to the cloud server database in a network mode.

The cloud server mainly has a database and an AI learning function, wherein the database stores the electricity consumption data uploaded by the edge end and is used for data access of an application end, and the AI learning function is used for training a V-I track diagram of the electric appliance to generate a neural network model and downloading the neural network model to an AI processor through wireless communication equipment.

Fig. 1 is the overall schematic diagram of the utility model, which shows the relationship between the non-invasive load recognition device and the appliance to be monitored, and the non-invasive load recognition device is installed at the entrance of the user's power consumption, and the switching condition of different appliances is recognized by monitoring the current, voltage and power value.

Fig. 2 is a structural view of a device for non-intrusive load recognition, which is mainly divided into three layers,

end/edge layer: the device comprises a sensor, an electric energy metering chip, an MCU, an AI module and a communication module;

cloud server: AI learning, database;

an application layer: and household electricity monitoring applications, such as WEB and mobile phone APP.

The specific realization principle is as follows:

the method comprises the following steps: load data collection

Load data acquisition is the basis of realizing load identification, through voltage transformer, current sampling circuit, power consumption data such as electric energy collection chip real-time supervision electric current, voltage, power, figure 3 shows identification device's structure chart, wherein the electric energy collection chip adopts the single-phase multi-functional measurement chip ATT7053 of taking SPI communication interface, 19bit Sigma-Delta ADC has, can be with electric current, voltage conversion digital quantity, still support simultaneously and calculate active power, reactive power on the piece, give main control STM32L476 with data transmission through the SPI interface and handle.

Step two: extracting current and voltage data of switching electrical appliance

By monitoring the power, when the power change is larger than a set threshold value, it is determined that an electrical appliance switching event occurs, and current and voltage data of the switching electrical appliance need to be further extracted for load identification.

1. Firstly, recording the switching time t of the electric appliance, which indicates that the state of some electric appliance is changed at the time t;

2. intercepting a current sequence { IIprev } and a voltage sequence { Vvprev } within T seconds before the time T;

3. intercepting a current sequence { IInext } and a voltage sequence { VNext } within T seconds after the T moment is stabilized;

4. carrying out Fourier transform on the { IIprev }, { Vvprev }, { IInext }, and { Vvenext }, taking a sampling point with the fundamental voltage phase angle closest to 0 as an initial sampling point of each period, and taking the average value of current and voltage of the same sampling point of N periods as a steady-state current and a voltage sequence before and after switching of an electric appliance, wherein the average value is the steady-state current { Iprev }, the steady-state voltage { Vprev } before switching, the steady-state current { Inext } after switching and the steady-state voltage { Vnext } after switching;

5. calculating the voltage of switching electrical appliance

Current { Ieq } - { | Iprev-Inext | }.

The embodiment of the utility model provides a V-I draws algorithm includes.

Step three: V-I trajectory graph synthesis

And (4) generating a V-I locus diagram for the current { Ieq } and the voltage { Veq } extracted in the step two.

Step four: V-I trace map identification

Passing the V-I image generated in the third step through I²The interface C is transmitted to an AI chip K210 to carry out image recognition of a neural network, a neural network model is trained and completed in a cloud server in advance and is downloaded to Flash of the K210 through a network, and only the model needs to be loaded during recognition. Meanwhile, a user can mark the V-I track map of the unknown electrical appliance, the cloud server can retrain the new V-I track map to generate a new neural network model, and then the model in the K210 Flash is updated through the network, so that the identification is more accurate.

Step five: recognition result uploading

The type and power of switching electrical appliances and the switching event time are uploaded to the cloud server database through the WIFI module RW007, the data of the cloud server database can be read by the household electricity monitoring application of the application layer, electricity utilization conditions are displayed for users, and the users can know the running states of the household electrical appliances in real time.

Fig. 3 is a structural diagram of an identification device according to an embodiment of the present invention.

The embodiment of the utility model provides an among the non-invasive load identification based on V-I orbit image identification, the principle as follows:

load data acquisition: and acquiring load data through the electric energy acquisition chip.

And extracting data of current { Ieq } and voltage { Veq } of the switching electrical appliance.

And (3) synthesizing a V-I locus diagram: and generating a V-I locus diagram for the current { Ieq } and the voltage { Veq } extracted in the S102.

And identifying the V-I track map: passing the V-I image generated in S103 through I²The interface C is transmitted to the AI chip K210 for image recognition of the neural network.

Uploading the recognition result: the type and power of switching electrical appliances and the switching event time are uploaded to a cloud server database through a WIFI module RW 007.

FIG. 4 is a diagram of a neural network model architecture for V-I trajectory graph recognition.

The above description is only for the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto, and any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be covered within the protection scope of the present invention by those skilled in the art within the technical scope of the present invention.

Claims (3)

1. A non-invasive load recognition apparatus based on trace image recognition, wherein the non-invasive load recognition apparatus based on trace image recognition comprises: the current transformer and the voltage transformer are respectively installed at the inlet of the household electrical appliance and are used for collecting current and voltage values of the household electrical appliance;

the electric energy metering chip converts analog quantity output by the current and voltage sensors into digital quantity, counts active power and reactive power, and transmits real-time current, voltage value and power value to the main control chip;

the main control chip is used for processing data generated by the current transformer, the voltage transformer and the electric energy metering chip, judging whether an electric appliance switching event occurs according to the power change condition, if the electric appliance switching occurs, generating a V-I track image of the switching electric appliance according to the current and voltage data, and transmitting the generated V-I track image to the AI processor;

AI processor by I²C, connecting the V-I locus diagram with a main control chip, and identifying the V-I locus diagram to obtain a result;

the wireless communication equipment is used for transmitting the identification result output by the AI processor to the cloud server database in a network mode;

and the cloud server stores and accesses the electricity utilization data uploaded by the main control chip through the database and is connected with the AI processor through the wireless communication equipment.

2. The apparatus for non-invasive load identification based on trace image recognition as claimed in claim 1, wherein the apparatus for non-invasive load identification based on trace image recognition further comprises:

the current sampling circuit monitors electricity consumption data of current, voltage and power in real time and the electric energy acquisition chip monitors the electricity consumption data of the current, the voltage and the power in real time;

the electric energy acquisition chip is provided with an SPI communication interface and transmits data to the main control chip through the SPI interface.

3. The apparatus for non-invasive load identification based on trace image recognition as claimed in claim 1, wherein the apparatus for non-invasive load identification based on trace image recognition further comprises:

and the smart phone is connected with the cloud server through wireless communication equipment.

Images