NON-INTRUSIVE LOAD MONITORING METHOD BASED ON V-I TRAJECTORY AND NEURAL NETWORK

Abstract

A non-intrusive load monitoring method based on V-I trajectory and neural network includes: collecting the household voltage, current and active power data in real time; determining whether there is a switching event and whether the load operating state has reached a steady state through the change of the active power; obtaining the voltage, current and power data of the load, converting the V-I trajectory into RGB color image containing the phase difference between the voltage and current, power and other information. After obtaining the RGB color image, performing normalization processing and performing load monitoring through pre-trained convolutional neural network. The present disclosure fully extracts the steady-state feature of the load through the convolutional neural network, and the neural network model can directly run on an embedded device, and does not need to rely on the computing support of a server.

Claims

1. A non-intrusive load monitoring method based on V-I trajectory and neural network, comprising: step 1, collecting voltage, current and power data of an electric household side in real time, and performing filtering; step 2, determining whether a switching event occurs through a bilateral sliding window algorithm, and when no switching event occurs, returning to step 1; step 3, when it is detected that a switching event occurs, obtaining, after a load reaches a steady state, voltage, current and power data of the load according to data in the steady state before and after the switching event; step 4, obtaining a V-I trajectory through the voltage and current data in the steady state obtained in step 3, and converting the V-I trajectory into an RGB image with a size of 2N*2N; wherein a power is expressed as a pixel value of the RGB image, and wherein said converting the V-I trajectory into the RGB image with the size of 2N*2N comprises: step 4.1, defining an initial value of each pixel to (0, 0, 0); step 4.2, obtaining, according to the obtained voltage and current at the steady state of the load, a maximum absolute value Umax of the voltage and a maximum absolute value Imax of the current; step 4.3, calculating Δu=Umax/N and Δi=Imax/N; step 4.4, calculating $\mathrm{Xj}=N+\mathrm{int}\left(\frac{\mathrm{Uj}}{\Delta u}\right),$ Yj=N+int(Ij/Δi), as RGB pixel coordinates to be defined, for each sampling point (Uj, Ij), where 0<j≤sample, and sample is a sampling point number in each cycle; and step 4.5, defining a pixel value correspondingly according to an active power value of the load; dividing the V-I trajectory into three stages, and then setting pixels in each stage to three different colors, comprising: If 0<j<200/3; a pixel value (Xj, Yj) is defined as (color_value, 0, 0); If 200/3<j<2*200/3; the pixel value (Xj, Yj) is defined as (0, color_value, 0); else; the pixel value (Xj, Yj) is defined as (0, 0, color_value); wherein color_value represents the defined pixel value; and step 5, normalizing the RGB image obtained in step 4, and using a monitoring network to obtain a load monitoring result; wherein the monitoring network comprises a convolutional neural network, which takes historical operation data of an electrical equipment and an RGB color image based on features of the V-I trajectory constructed by the historical operation data as truth values for training.

2. The non-intrusive load monitoring method based on V-I trajectory and neural network according to claim 1, wherein the monitoring network comprises two convolution layers, two pooling layers and three fully connected layers, and runs on a MCU of STM32F7, or runs on a computer or a server by using Alexnet model.

3. The non-intrusive load monitoring method based on V-I trajectory and neural network according to claim 1, wherein in step 2, said determining whether the switching event occurs comprises: step 2.1, setting two sliding windows, and removing a maximum and a minimum values in each window; and step 2.2, calculating a difference between mean values of the two windows, and if the difference is greater than a preset threshold, determining that the switching event has occurred.

4. The non-intrusive load monitoring method based on V-I trajectory and neural network according to claim 1, wherein in step 4, the Umax and Imax of a high-power load are defined as fixed values which are greater than the Umax or Imax of the high-power load.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0026] FIG. 1 is a flow chart of the method according to the present disclosure;

[0027] FIG. 2 is a schematic structural diagram of a convolutional neural network model according to an embodiment of the present disclosure;

[0028] FIG. 3 is a flow chart of the bilateral sliding window algorithm;

[0029] FIG. 4 is a feature diagram of some loads (left: air conditioner, middle: refrigerator, right: electric lamp) based on the V-I trajectory according to an embodiment of the present disclosure;

[0030] FIG. 5 is a gray diagram showing R, G and B channels, respectively, in the air conditioner feature diagram (left: R channel, middle: G channel, right: B channel).

DESCRIPTION OF EMBODIMENTS

[0031] The present disclosure is illustrated with reference to the drawings and the implementation using the BLUED public dataset, and the specific implementation step are as follows:

[0032] The present disclosure provides a non-intrusive load monitoring method based on V-I trajectory and neural network, as shown in FIG. 1, the implementation steps of the method include: [0033] S1: firstly, extracting the data of 5 kinds of household electrical equipment from the BLUED dataset, and then constructing an RGB color diagram based on the V-I trajectory features, and training a convolutional neural network model as a monitoring network, such as the Alexnet model. The monitoring network model in the embodiment is shown in the figures. In order to run directly on MCUs above STM32F7, the constructed monitoring network model includes two convolution layers, two pooling layers and three fully connected layers, which is not complicated. The specific structure is shown in FIG. 2. [0034] S2: collecting voltage, current and power data of an electric household side in real time, and performing filtering processing to the obtained voltage and current data. In the embodiment, the voltage and current sampling frequency of the BLUED public dataset is 12 KHz, the power value frequency is 60 Hz, and each cycle includes 200 sampling points. [0035] S3: determining whether a switching event occurs through the bilateral sliding window algorithm. In the embodiment, the specific parameters are as follows: setting two sliding windows with a window size of (5, 5), and removing the maximum and minimum values in each windows, calculating the mean value of each window, calculating the difference between the mean values, then comparing the difference with a preset threshold, and determining that the switching event has occurred, if the difference between the mean values is greater than the preset threshold, of which the process is shown in FIG. 3. [0036] S4: after the load reaches the steady state for three times consecutively, obtaining the voltage, current and power data of the load at the steady state. [0037] S5: obtaining the V-I trajectory according to the steady-state voltage and current data obtained in S4, and then converting the V-I trajectory into an RGB image with a size of 2N*2N. In the embodiment, N is 32; the specific steps are as follows: [0038] (1) Firstly, defining an initial value of each pixel as (0, 0, 0). [0039] (2) Calculating the maximum absolute values Umax and Imax of the voltage and current of the low-power load, and directly defining the Imax of the high-power load as a fixed value, so that the V-I trajectory can contain all the current information. In the embodiment, the loads with the active power values of less than 510 W are regarded as low-power loads, and the others are regarded as high-power loads. For the high-power loads, defining the fixed value of Umax as 400V and the fixed value of Imax s 20 A. In this way, the V-I trajectory may contain most of the current data. [0040] (3) Calculating Δu=Umax/N and Δi=Imax/N. [0041] (4) Calculating

[00002]$\mathrm{Xj}=N+\mathrm{int}\left(\frac{\mathrm{Uj}}{\Delta u}\right),$

Yj=N+int(Ij/Δi) for each sampling point (Uj, Ij) (0<j≤200) without the continuous processing of the trajectory. [0042] (5) Defining the corresponding pixel value according to the active power of the load. When the active power P is greater than 510 W, the electrical equipment with a relatively large power value has obvious features, and thus can also be correctly monitored through general V-I trajectory features. Therefore, the value of each pixel point is directly set color_value=255. When the active power P is less than 510 W, defining color_value=int(P/2). [0043] (6) In order to fully reflect the steady-state features of the load in the RGB image, the specific process of defining the pixel value (Xj, Yj) is as follows:

[0044] If 0<j<200/3:

[0045] The pixel value (Xj, Yj) is defined as (color_value, 0, 0).

[0046] If 200/3<j<2*200/3:

[0047] The pixel value (Xj, Yj) is defined as (0, color_value, 0).

[0048] else:

[0049] The pixel value (Xj, Yj) is defined as (0, 0, color_value).

[0050] The V-I trajectory feature diagram obtained in this way can reflect the load features such as the phase difference of the voltage and current, impedance features and power. FIG. 4 below is a trajectory features diagram of some loads in the embodiment. In FIG. 4, the high-power load and low-power load can be seen directly by the naked eye, in which the power of the electric lamp is smaller, so the corresponding brightness is smaller. Because the Umax of the high-power load is directly defined as 400V, the trajectory feature diagram is relatively concentrated in the middle, and the feature diagram of the low-power load covers the entire area. Then each load trajectory feature diagram is composed of three colors of red, green and blue (the three channels R, G and B of the trajectory in the air conditioner feature diagram in FIG. 5 are shown separately), and has directions. The brightness of the low-power load feature diagram is also various according to the power.

[0051] S6: normalizing the RGB image obtained in S5, inputting it into the pre-trained convolutional neural network, and obtaining the monitoring result. Because the input end of the neural network is a picture, the normalization processing in the embodiment is very simple, which just divides the value of each pixel point directly by 255. The RGB image in the present disclosure already contains information such as the V-I trajectory features, phase difference between the voltage and current, active power, etc., so the monitoring effect is much better than the method by using the V-I trajectory or power information alone. Moreover, the convolutional neural network used is not complicated, and thus can be directly run on embedded devices, and further improve the real-time performance, and does not rely on the computing support of the servers.

[0052] In this application, the term “controller” and/or “module” may refer to, be part of, or include: an Application Specific Integrated Circuit (ASIC); a digital, analog, or mixed analog/digital discrete circuit; a digital, analog, or mixed analog/digital integrated circuit; a combinational logic circuit; a field programmable gate array (FPGA); a processor circuit (shared, dedicated, or group) that executes code; a memory circuit (shared, dedicated, or group) that stores code executed by the processor circuit; other suitable hardware components (e.g., op amp circuit integrator as part of the heat flux data module) that provide the described functionality; or a combination of some or all of the above, such as in a system-on-chip.

[0053] The term memory is a subset of the term computer-readable medium. The term computer-readable medium, as used herein, does not encompass transitory electrical or electromagnetic signals propagating through a medium (such as on a carrier wave); the term computer-readable medium may therefore be considered tangible and non-transitory. Non-limiting examples of a non-transitory, tangible computer-readable medium are nonvolatile memory circuits (such as a flash memory circuit, an erasable programmable read-only memory circuit, or a mask read-only circuit), volatile memory circuits (such as a static random access memory circuit or a dynamic random access memory circuit), magnetic storage media (such as an analog or digital magnetic tape or a hard disk drive), and optical storage media (such as a CD, a DVD, or a Blu-ray Disc).

[0054] The apparatuses and methods described in this application may be partially or fully implemented by a special purpose computer created by configuring a general-purpose computer to execute one or more particular functions embodied in computer programs. The functional blocks, flowchart components, and other elements described above serve as software specifications, which can be translated into the computer programs by the routine work of a skilled technician or programmer.

[0055] The steps of the method or algorithm described combined with the embodiments of the present disclosure may be implemented in a hardware manner, or may be implemented in a manner in which a processor executes software instructions. The software instructions may consist of corresponding software modules, and the software modules can be stored in Random Access Memory (RAM), flash memory, Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), registers, hard disks, removable hard disks, CD-ROMs or any other forms of storage media well-known in the art. An exemplary storage medium is coupled to the processor, such that the processor can read information from, and write information to, the storage medium. The storage medium can also be an integral part of the processor. The processor and storage medium may reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the ASIC may be located in a node device, such as the processing node described above. In addition, the processor and storage medium may also exist in the node device as discrete components.

[0056] It should be noted that when the data compression apparatus provided in the foregoing embodiment performs data compression, division into the foregoing functional modules is used only as an example for description. In an actual application, the foregoing functions can be allocated to and implemented by different functional modules based on a requirement, that is, an inner structure of the apparatus is divided into different functional modules, to implement all or some of the functions described above. For details about a specific implementation process, refer to the method embodiment. Details are not described herein again.

[0057] All or some of the foregoing embodiments may be implemented by using software, hardware, firmware, or any combination thereof. When the software is used for implementation, all or some of the embodiments may be implemented in a form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a server or a terminal, all or some of the procedures or functions according to the embodiments of this application are generated. The computer instructions may be stored in a computer-readable storage medium or may be transmitted from a computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, computer, server, or data center to another website, computer, server, or data center in a wired (for example, a coaxial optical cable, an optical fiber, or a digital subscriber line) or wireless (for example, infrared, radio, or microwave) manner. The computer-readable storage medium may be any usable medium accessible by a server or a terminal, or a data storage device, such as a server or a data center, integrating one or more usable media. The usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, a digital video disk (DVD)), or a semiconductor medium (for example, a solid-state drive).

[0058] Obviously, the above mentioned embodiments are only examples for clear description, rather than limiting the implementation. For those skilled in the art, on the basis of the above description, other different forms of changes or variations can also be made. It is unnecessary and impossible to exhaust all implementations here. However, the obvious changes or variations derived therefrom are still within the protection scope of the present disclosure.