A non-intrusive load identification method and device based on PLC communication

Abstract

This invention provides a non-intrusive load identification method and apparatus based on PLC communication. It includes a preset feature library and real-time input data for detection; buffering 10-20 cycles of real-time data; performing detection and calculation on the real-time data using power and current fluctuation methods; determining whether a switching event has occurred based on the calculation results of the power or current fluctuation methods; if yes, proceeding to the load identification step; if no, continuing detection and calculation; after entering the load identification step, buffering 4-5 cycles of real-time data; extracting features from this set of real-time data to obtain the current event feature; performing identification calculation based on the current event feature and the feature library; and outputting the load category based on the identification calculation result. This invention can accurately determine the occurrence of switching events and precisely determine the type of load feature by calculating and extracting four features within a certain period, thus improving the accuracy of identification.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution safety monitoring and control technology, and in particular to a non-intrusive load identification method and device based on PLC communication. Background Technology

[0002] Non-intrusive load identification refers to installing equipment at the power inlet and analyzing signals such as voltage and current to determine the type and operating status of individual loads within a load cluster. This method is frequently mentioned and applied. Conventional load type identification methods employ static identification strategies based on single characteristic groups to improve accuracy. Furthermore, to suppress inrush current, the inrush current generated by the connection or disconnection of electrical loads can be used to stabilize the current.

[0003] However, conventional identification methods cannot guarantee the accuracy of identifying switching events and load types, resulting in low identification speed and accuracy, and difficulties in monitoring switching status. Load type identification plays an important role in power systems. Summary of the Invention

[0004] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of this invention is to provide a non-intrusive load identification method and apparatus based on PLC communication to solve the above problems.

[0005] The technical solution adopted by this invention to solve its technical problem is:

[0006] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of this invention is to provide a non-intrusive load identification method and apparatus based on PLC communication to solve the above problems.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] A non-intrusive load identification method based on PLC communication, characterized in that it includes a preset feature library and further includes:

[0009] Detect real-time input data;

[0010] Cache real-time data for 10-20 cycles;

[0011] Real-time data is detected and calculated using the power method and the current fluctuation method; the result of the power method or the current fluctuation method is used to determine whether a switching event has occurred.

[0012] If yes, proceed to the load identification step; if no, continue with detection and calculation.

[0013] After entering the load identification step, cache 4-5 cycles of real-time data first;

[0014] Extract the Fourier features, wavelet features, statistical features, and VI curve features from this set of real-time data to obtain the current event features;

[0015] Identification calculations are performed based on the current event characteristics and the feature library.

[0016] The load category is output based on the results of the identification calculation; the real-time data includes current i, voltage v, and power P.

[0017] As a further improvement of the present invention: A preset power threshold is used, and the calculation method for detecting and calculating real-time data using the power method is as follows:

[0018] Compare the power P of two adjacent cycles;

[0019] If |P current -P previous |>P threshold Among them, P current For the current cycle power, P previous P is the power of the previous cycle. threshold If the threshold value is reached, it is considered that a switching event has occurred in the current cycle; otherwise, it is considered that no switching event has occurred in the current cycle, and the detection continues.

[0020] As a further improvement of the present invention: A preset current threshold is used, and the calculation method for detecting and calculating real-time data using the current fluctuation method is as follows: The maximum current value in each cycle is linearly interpolated to achieve an effective current packet. That is, calculate:

[0021]

[0022] Where x k =(k-id) max-next ) / (id max-last -id max-nex );

[0023] Among them, i max-last and ID max-last It is the value and position of a current maximum at the interpolation position; i max-next and ID max-next It is the value and location of the next maximum current;

[0024] Then calculate: in, The average value of the effective current packets is calculated over two period-length windows; T is the number of samples per period; L = 2T, where L is the window length; finally, the variance is calculated. k∈i; n∈i;

[0025] If the variance is greater than the current threshold, it is considered that a switching event has occurred in the current cycle; otherwise, it is considered that no switching event has occurred in the current cycle, and the detection continues.

[0026] As a further improvement of the present invention: the method for calculating the Fourier features is as follows: C n =∑ k i k e jkn / N , where C n =∑ k i k e jkn / N , of which F fft，m The features are Fourier features, where N is the total number of input data points and j is the imaginary identifier. m∈N; n∈N; e is a mathematical constant.

[0027] As a further improvement of the present invention: the wavelet feature is calculated using a 5th-order wavelet transform, with the formula for calculating 5 iterations as follows: Where g is the high-pass filter and h is the low-pass filter coefficient; the transformed wavelet coefficients are [D1, D2, D3, D4, D5, A5];

[0028] Formula F for calculating wavelet transform features dwt，m =mean(D m [n] 2 Wavelet features are obtained.

[0029] As a further improvement of the present invention, the statistical characteristics are calculated by calculating the maximum value and standard deviation of the current i and power P in the current cycle, respectively, to obtain the statistical characteristics.

[0030] As a further improvement of the present invention: the VI curve features include the length of the VI curve, the enclosing area, the degree of asymmetry between the positive and negative sides, and the distance between the highest and lowest points.

[0031] As a further improvement of the present invention: the identification calculation method of "performing identification calculation based on the output current event features and feature library" includes the k-nearest neighbor method and the decision tree method;

[0032] The k-nearest neighbor method determines the load category by comparing the features of the load to be identified with the feature library. Among them, the top k min values ​​are the k largest values;

[0033] The decision tree method collects training samples, segments the dataset based on maximum information entropy gain, iteratively generates decision trees, obtains a trained decision tree model, and then inputs the features to be identified into the decision tree model to obtain the load category.

[0034] A non-intrusive load identification device based on PLC communication, wherein the non-intrusive load identification method based on PLC communication described in any one of the above-mentioned methods further includes:

[0035] Signal acquisition and processing unit: used to acquire and buffer data, and then convert the data into signals and send them to the central processing unit;

[0036] Central Processing Unit: Used to receive signals from the signal acquisition and processing unit, and to process the information carried by the signals.

[0037] The data is identified, calculated, judged, and its features are extracted.

[0038] Terminal access unit: used to connect the identification device to the terminal equipment;

[0039] Communication link unit: used to connect devices to a network for communication;

[0040] Human-computer interaction unit: used to view results, restore factory settings, or reset and clear cache;

[0041] Interface unit: used to connect the antenna;

[0042] Circuit matching unit: used to perform voltage division and coupling of the connected circuits to match the central processing unit;

[0043] The interface unit is connected to the central processing unit through the communication link unit, the circuit matching unit is connected to the signal acquisition and processing unit, and the signal acquisition and processing unit, the human-computer interaction unit, and the terminal access unit are connected to the central processing unit.

[0044] As a further improvement of the present invention: the central processing unit includes a communication link management module, a feature recognition module, a feature extraction module, a switching detection module, a terminal access management module, and a central ARM processor; the communication link management module, the feature recognition module, the switching detection module, and the terminal access management module are connected to the central ARM processor, and the feature extraction module is connected to the feature recognition module.

[0045] As a further improvement of the present invention: the signal acquisition and processing unit includes a phase line voltage analog-to-digital conversion module, a phase line current analog-to-digital conversion module, a neutral line current analog-to-digital conversion module, and a digital signal processing module; the phase line voltage analog-to-digital conversion module, the phase line current analog-to-digital conversion module, and the neutral line current analog-to-digital conversion module are connected to the digital signal processing module.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] By using two detection methods simultaneously in the central processing unit to detect whether a switching event has occurred, the occurrence of a switching event can be accurately determined. Furthermore, by calculating and extracting four features within a certain period, the type of load feature can be accurately determined, thereby improving the accuracy of recognition. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the process of the present invention.

[0049] Figure 2 This is a schematic diagram showing the connection relationship between the various units of the present invention.

[0050] Figure 3 This is an exploded view of the device structure in Embodiment 2 of the present invention.

[0051] Figure 4 This is a diagram of an apparatus according to Embodiment 2 of the present invention.

[0052] Figure label explanations: 01-Central Processing Unit; 101-Central ARM Processor; 102-Communication Link Management Module; 103-Terminal Access Management Module; 104-Feature Recognition Module; 105-Feature Extraction Module; 106-Switching Detection Module; 02-Signal Acquisition and Processing Unit; 201-Phase Line Voltage Analog-to-Digital Conversion Module; 202-Phase Line Current Analog-to-Digital Conversion Module; 203-Neutral Line Current Analog-to-Digital Conversion Module; 204-Digital Signal Processing Module; 03-Terminal Access Unit; 301-HPLC Power Carrier communication CCO module; 302-RS-485 serial communication module; 04-Communication link unit; 401-4G module; 402-Ethernet module; 403-Wi-Fi module; 05-Human machine interaction unit; 501-Button interaction module; 502-LCD display interaction module; 06-Interface unit; 601-Live wire interface; 602-Neutral wire interface; 603-Antenna interface; 604-Ethernet interface; 605-RS-485 interface; 07-Circuit matching unit; 701-Electrical coupling module. Detailed Implementation

[0053] The present invention will now be further described in conjunction with the accompanying drawings and embodiments:

[0054] Example 1:

[0055] This embodiment provides the following: Figure 1 The non-intrusive load identification method based on PLC communication shown is characterized by including a preset feature library and further comprising:

[0056] Detect real-time input data;

[0057] Cache real-time data for 10-20 cycles;

[0058] Real-time data is detected and calculated using the power method and the current fluctuation method; the result of the power method or the current fluctuation method is used to determine whether a switching event has occurred.

[0059] If yes, proceed to the load identification step; if no, continue with detection and calculation.

[0060] After entering the load identification step, cache 4-5 cycles of real-time data first;

[0061] Extract the Fourier features, wavelet features, statistical features, and VI curve features from this set of real-time data to obtain the current event features;

[0062] Identification calculations are performed based on the current event characteristics and the feature library.

[0063] The load category is output based on the results of the identification calculation; the real-time data includes current i, voltage v, and power P.

[0064] In this first embodiment, data of current i, voltage v, and power P for 10-20 cycles are cached for detecting switching events. Each current i, voltage v, and power P contains multiple values. To ensure the real-time performance and accuracy of the switching algorithm, two fast switching identification algorithms, namely the power method and the current fluctuation method, are used for identification calculation to determine whether a switching event has occurred in that cycle.

[0065] The preset power threshold and the calculation method for real-time data detection using the power method are as follows:

[0066] Compare the power P of two adjacent cycles;

[0067] If |P current -P previous |>P threshold Among them, P current For the current cycle power, P previous P is the power of the previous cycle. thresjold If the power threshold is met, a switching event is considered to have occurred in the current cycle; otherwise, no switching event is considered to have occurred in the current cycle, and the detection continues. In other words, the power of two adjacent cycles is compared; if the power difference is greater than the power threshold, a switching event is considered to have occurred. The power threshold can be modified.

[0068] Furthermore, a preset current threshold is established, and the calculation method for detecting and calculating real-time data using the current fluctuation method is set as follows: linear interpolation is performed on the maximum current value of each cycle to achieve an effective current packet.

[0069] That is, calculate:

[0070]

[0071] Where x k =(k-id) max-next ) / (id max-last -id max-nex );

[0072] Among them, i max-last and ID max-last It is the value and position of a current maximum at the interpolation position; i max-next and ID max-next It is the value and location of the next maximum current;

[0073] Then calculate: in, The average value of the effective current packets is calculated over two period-length windows; T is the number of samples per period; L = 2T, where L is the window length; finally, the variance is calculated. k∈i; n∈i; that is, k and n are the kth or nth numbers in the current i data.

[0074] If the variance is greater than the current threshold, a switching event is considered to have occurred in the current cycle; otherwise, no switching event is considered to have occurred in the current cycle, and the detection continues. The current threshold can be modified.

[0075] It should be noted that the power method and the current fluctuation method detection algorithms are executed in parallel, that is, they are performed simultaneously. Once either detection algorithm identifies a switching event, it immediately enters the load identification step.

[0076] In the load identification step of this embodiment, the current i, voltage v, and power P data for 4-5 cycles before and after the switching event are first cached, and then feature extraction and load identification are performed. Feature extraction includes time-frequency domain features from Fourier transform and wavelet transform, statistical features, and VI curve features.

[0077] The Fourier transform features are primarily applied to current data, and the calculation method for these Fourier features is as follows: C n =∑ k i k e jkn / N , where C n =∑ k i k e jkn / N , of which F fft，m The features are Fourier features, where N is the total number of input data points and j is the imaginary identifier. m∈N; n∈N; that is, m and n are both numbers in N; k is the kth number in the current i data; e is a mathematical constant.

[0078] Specifically, a 5th-order wavelet transform is used in the wavelet transform. The calculation method for the wavelet features is as follows: using a 5th-order wavelet transform, for an input A0 = i, the formula for calculating 5 iterations is: Where g is the high-pass filter and h is the low-pass filter coefficient; the transformed wavelet coefficients are [D1, D2, D3, D4, D5, A5];

[0079] Formula F for calculating wavelet transform features dwt，m =mean(D m [n] 2 The wavelet features are obtained by averaging the squares of each term of the wavelet coefficients.

[0080] The statistical characteristics are calculated by calculating the maximum value and mean square error of the current i and power P in the current cycle, respectively, to obtain the statistical characteristics.

[0081] The VI curve features include the length of the VI curve, the area enclosed, the degree of asymmetry between the positive and negative sides, and the distance between the highest and lowest points.

[0082] In this embodiment, the identification calculation method of "performing identification calculation based on the output current event features and feature library" includes the k-nearest neighbor method and the decision tree method;

[0083] The k-nearest neighbor method determines the load category by comparing the features of the load to be identified with the feature library. Where, the top k min values ​​are taken as the k smallest values; the specific meaning of this formula is that, firstly, through |f 待识别 -f g |2 Calculate the distance between the load feature to be identified and the load features in the feature library, then select the group of features with the smallest distance in the feature library, select the type with the most types in the group of features, obtain the load feature type of the load feature to be identified, and output the type.

[0084] The decision tree method involves collecting training samples, segmenting the dataset based on maximum information entropy gain, iteratively generating decision trees to obtain a trained decision tree model, and then inputting the features to be identified into the decision tree model to obtain the load type.

[0085] Example 2:

[0086] This embodiment provides the following: Figure 2-4 The non-intrusive load identification device based on PLC communication shown herein, wherein the non-intrusive load identification method based on PLC communication described in any one of the above-mentioned methods further includes:

[0087] Signal acquisition and processing unit 02: used to acquire and buffer data, and convert the data into signals before sending them to central processing unit 01;

[0088] Central processing unit 01: Used to receive signals from signal acquisition and processing unit 02, and to identify, calculate, judge and extract features of the data carried by the signals;

[0089] Terminal access unit 03: used to connect the identification device to the terminal equipment;

[0090] Communication link unit 04: used to connect the device to a network for communication;

[0091] Human-computer interaction unit 05: Used to view results, restore factory settings, or reset and clear the cache;

[0092] Interface unit 06: Used for connecting an antenna;

[0093] Circuit matching unit 07: Used to perform voltage division and coupling of the connected circuit to match the central processing unit 01.

[0094] As attached Figure 2 As shown, the interface unit 06 is connected to the central processing unit 01 via the communication link unit 04, the circuit matching unit 07 is connected to the signal acquisition and processing unit 02, and the signal acquisition and processing unit 02, the human-computer interaction unit 05, and the terminal access unit 03 are connected to the central processing unit 01. (See attached diagram) Figure 3 As shown, the human-machine interaction unit 05 is located at the top of the device. Below the human-machine interaction unit 05, the central processing unit 01 and the terminal access unit 03 are located. The two units are located on the same mounting plate. The signal acquisition and processing unit 02 is located below the terminal access unit 03, and the communication link unit 04 is located below the central processing unit 01. The above arrangement facilitates the connection between lines.

[0095] Furthermore, the central processing unit 01 includes a communication link management module 102, a feature recognition module 104, a feature extraction module 105, a switching detection module 106, a terminal access management module 103, and a central ARM processor 101; the communication link management module 102, the feature recognition module 104, the switching detection module 106, and the terminal access management module 103 are connected to the central ARM processor 101, and the feature extraction module 105 is connected to the feature recognition module 104.

[0096] Furthermore, the signal acquisition and processing unit 02 includes a phase line voltage analog-to-digital converter module 201, a phase line current analog-to-digital converter module 202, a neutral line current analog-to-digital converter module 203, and a digital signal processing module 204; the phase line voltage analog-to-digital converter module 201, the phase line current analog-to-digital converter module 202, and the neutral line current analog-to-digital converter module 203 are connected to the digital signal processing module 204. The signal acquisition and processing unit serves as the front-end for electrical signal acquisition and processing, outputting the processed signal to the central control unit. The signal acquisition and processing unit 02 employs three high-precision analog-to-digital converter modules to digitize the live wire voltage, live wire current, and neutral line current from analog signals, respectively; the digital signal processing module 204 employs a dedicated DSP module to normalize, filter, perform time-domain conversion, and frequency-domain conversion on the digitized live wire voltage, live wire current, and neutral line current.

[0097] Furthermore, the human-machine interaction unit 05 includes a button interaction module 501 and an LCD display interaction module 502. The button interaction module 501 has two buttons: one for restoring the entire recognition device to factory settings, and the other for resetting only, preserving the contents of the feature database but clearing the cached real-time data. The LCD display interaction module 502 displays the results, including electrical parameters and alarm information. The human-machine interaction unit 05 provides input of control commands and display of results. (See attached...) Figure 4 As shown, both the button interaction module 501 and the LCD display interaction module 502 are located on the top surface of the device.

[0098] Furthermore, the terminal access unit 03 includes an HPLC power line carrier communication CCO module 301 and an RS-485 serial communication module 302. The terminal access unit 03 provides protocol conversion between the electrical terminal and the central control unit, adapting to different types of access protocols. The system connects to terminal electrical devices, such as batteries and chargers, via both HPLC and RS-485 communication methods.

[0099] Furthermore, the circuit matching unit 07 includes a circuit coupling module 701.

[0100] Furthermore, the communication link unit 04 includes a 4G module 401, an Ethernet module 402, and a WiFi module 403. The link standard can be selected according to requirements, or it can be selected adaptively.

[0101] Further details are attached. Figure 4As shown in the attached diagram, interface unit 06 includes a live wire interface 601, a neutral wire interface 602, an antenna interface 603, an Ethernet interface 604, and an RS-485 interface 605. Communication link unit 04 is the gateway's external link center, providing adaptation and selection for different communication standards. As shown in the attached diagram, the interfaces are distributed on both sides of the device, with the live wire interface 601, neutral wire interface 602, and RS-485 interface 605 located on one side, and the antenna interface 603 and Ethernet interface 604 located on the other side.

[0102] In this embodiment, the system acquires electrical signals from the access device via an ADC, performs digital filtering and digital signal processing by a DSP (Digital Signal Processing Module 204), and then sends the data to the central control unit. The central control unit performs electrical feature extraction, feature detection, and feature recognition, and then transmits the data to the cloud via 4G, Wi-Fi, or Ethernet for decision-making and control.

[0103] In this embodiment, as shown in the appendix Figure 3-4 As shown, the identification device is a load identification safety power gateway, and the gateway is designed for DIN rail mounting. It can be adapted to standard DIN rails of different depths, ranging from 35mm to 1mm (width to thickness).

[0104] Detailed description of this embodiment:

[0105] Specifically, it is a non-intrusive load identification safety power gateway based on power line carrier PLC communication, including a central control unit, a signal acquisition and processing unit 02, a communication link unit 04, a terminal access unit 03, and a human-machine interaction unit 05;

[0106] The signal acquisition and processing unit 02 includes a three-channel ADC analog-to-digital converter circuit and a DSP digital signal processing module 204. The sampled analog current and voltage signals are converted into discrete digital signals by the ADC, and then the digital signals are normalized and output to the DSP for processing. The DSP obtains the normalized data and further performs Fast Fourier Transform and Wavelet Transform on the data to obtain the frequency domain information, time-frequency domain information, and VI trajectory information of the voltage and current signals, which are then output to the central control unit.

[0107] The central control unit includes a central ARM processor 101, a terminal access management module 103, a communication link management module 102, a signal feature extraction module 105, a switching detection module 106, and a feature recognition module.

[0108] After the DSP outputs time-domain information, the ARM performs switching event detection. Upon detecting a switching change, the DSP is notified to perform Fourier and wavelet transforms to obtain frequency-domain information, time-frequency domain information, and VI trajectory information. The signal is then sent to the feature extraction module 105 to extract signal features. These features are further sent to the feature recognition module for load identification and fault arc identification, and the event type is output to the ARM.

[0109] Furthermore, the ARM uploads the event information to the external platform through the signal link module according to the output event type, and at the same time controls the power consumption behavior of the terminal through the terminal access management module 103 according to the safe power consumption control strategy.

[0110] The terminal access unit 03 includes an HPLC power line carrier communication CCO module 301 and an RS-485 serial communication module 302. The terminal device connects to the HPLC power line carrier communication CCO module 301 via a power line. The CCO can control the terminal device, and the terminal device can also send real-time status information to the CCO. The terminal device connects to the RS-485 serial communication module 302 via an RS-485 bus. The RS-485 serial communication module 302 can send control commands to the terminal, and the terminal can also send real-time status information to the RS-485 serial communication module 302.

[0111] The main function of this invention is to provide a method and apparatus for detecting various switching events and load types.

[0112] In summary, after reading this invention document, those skilled in the art can make various other corresponding modifications to the technical solutions and concepts based on this invention without creative mental effort, and all of these modifications fall within the scope of protection of this invention.

Claims

1. A non-intrusive load identification method based on PLC communication, characterized in that, Including a preset feature library, it also includes: Detect real-time input data; Cache real-time data for 10-20 cycles; Real-time data is detected and calculated using the power method and the current fluctuation method; the result of the power method or the current fluctuation method is used to determine whether a switching event has occurred. If yes, proceed to the load identification step; if no, continue with detection and calculation. After entering the load identification step, cache 4-5 cycles of real-time data first; Extract the Fourier features, wavelet features, statistical features, and VI curve features from this set of real-time data to obtain the current event features; Identification calculations are performed based on the current event characteristics and the feature library. The load category is output based on the results of the identification calculation; the real-time data includes current i, voltage v, and power P; The preset power threshold, and the calculation method for real-time data detection using the power method, are as follows: Compare the power P of two adjacent cycles; If |P current -P previous |>P threshold Among them, P current For the current cycle power, P previous P is the power of the previous cycle. threshold If the power threshold is met, a switching event is considered to have occurred in the current cycle; otherwise, no switching event is considered to have occurred in the current cycle, and the detection continues. The current fluctuation method, which uses a preset current threshold, performs real-time data detection and calculation as follows: It linearly interpolates the maximum current value for each cycle to achieve an effective current packet. That is, calculate: Where x k =(k-id) max-next ) / (id max-last -id max-next ); Among them, i max-last and ID max-last It is the value and position of a current maximum at the interpolation position; i max-next and ID max-next It is the value and location of the next maximum current; Then calculate: in, The average value of the effective current packets is calculated over two period-length windows; T is the number of samples per period; L = 2T, where L is the window length; finally, the variance is calculated. k∈i; n∈i; If the variance is greater than the current threshold, it is considered that a switching event has occurred in the current cycle; otherwise, it is considered that no switching event has occurred in the current cycle, and the detection continues. The identification calculation method described in "recognition calculation based on the current event features and feature library" includes the k-nearest neighbor method and the decision tree method; The k-nearest neighbor method determines the load category by comparing the features of the load to be identified with the feature library. Pred(f 待识别 = Most categories (top k min{f g ∈Feature library}(|f 识别库 -f g |2)), where the top k min values ​​are the k largest values; The decision tree method collects training samples, segments the dataset based on maximum information entropy gain, iteratively generates decision trees, obtains a trained decision tree model, and then inputs the features to be identified into the decision tree model to obtain the load category.

2. The non-intrusive load identification method based on PLC communication according to claim 1, characterized in that, The Fourier features are calculated as follows: Among them, c n =∑ k i k e jkn / N Among them, F fft，m The features are Fourier features, where N is the total number of input data points and j is the imaginary identifier. m∈N; n∈N; e is a mathematical constant.

3. The non-intrusive load identification method based on PLC communication according to claim 1, characterized in that, The wavelet features are calculated using a 5th-order wavelet transform, with the formula for 5 iterations as follows: Where g is the high-pass filter and h is the low-pass filter coefficient; the transformed wavelet coefficients are [D1, D2, D3, D4, D5, A5]; Formula F for calculating wavelet transform features dwt，m =mean(D m [n] 2 Wavelet features are obtained.

4. The non-intrusive load identification method based on PLC communication according to claim 1, characterized in that, The statistical characteristics are calculated as follows: the maximum value and mean square error of the current i and power P in the current cycle are calculated respectively to obtain the statistical characteristics.

5. The non-intrusive load identification method based on PLC communication according to claim 1, characterized in that, The VI curve features include the length of the VI curve, the area enclosed, the degree of asymmetry between the positive and negative sides, and the distance between the highest and lowest points.

6. A non-intrusive load identification device based on PLC communication, characterized in that, The non-intrusive load identification method based on PLC communication according to any one of claims 1-5 further includes: Signal acquisition and processing unit: used to acquire and buffer data, and then convert the data into signals and send them to the central processing unit; Central processing unit: Used to receive signals from the signal acquisition and processing unit, and to identify, calculate, judge and extract features from the data carried by the signals; Terminal access unit: used to connect the identification device to the terminal equipment; Communication link unit: used to connect devices to a network for communication; Human-computer interaction unit: used to view results, restore factory settings, or reset and clear cache; Interface unit: used to connect the antenna; Circuit matching unit: used to perform voltage division and coupling of the connected circuits to match the central processing unit; The interface unit is connected to the central processing unit through the communication link unit, the circuit matching unit is connected to the signal acquisition and processing unit, and the signal acquisition and processing unit, the human-computer interaction unit, and the terminal access unit are connected to the central processing unit.

7. A non-intrusive load identification device based on PLC communication according to claim 6, characterized in that, The central processing unit includes a communication link management module, a feature recognition module, a feature extraction module, a switching detection module, a terminal access management module, and a central ARM processor; the communication link management module, the feature recognition module, the switching detection module, and the terminal access management module are connected to the central ARM processor, and the feature extraction module is connected to the feature recognition module; The signal acquisition and processing unit includes a phase line voltage analog-to-digital conversion module, a phase line current analog-to-digital conversion module, and a neutral wire.

Images