A load identification method and system based on comprehensive multi-element information similarity analysis

Abstract

This invention discloses a load identification method and system based on comprehensive multivariate information similarity analysis, comprising: determining whether a load change exists based on the fundamental amplitude values ​​corresponding to the current data of the current window and the previous data window; when a load change is determined to exist and the change type is load shedding, acquiring the current data of the next data window; performing load decomposition based on the current data of the previous data window and the current data of the next data window to determine first load decomposition data; performing multivariate information analysis to determine a first multivariate feature information vector corresponding to the first load decomposition data and a second multivariate feature information vector corresponding to each second load decomposition data; determining third load decomposition data corresponding to the shedding load based on the first multivariate feature information vector and the second multivariate feature information vector; and performing load matching based on the correlation between the already engaged load and the second load decomposition data according to the third load decomposition data to determine the shedding load.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution technology, and more specifically, to a load identification method and system based on comprehensive multi-source information similarity analysis. Background Technology

[0002] With the continuous increase in global energy consumption in recent years, the problem of energy shortage has become increasingly prominent. Energy transformation through technological means is conducive to the shift of the energy system towards low-carbon and green development. Currently, the connection between customer-side resources such as power distribution and consumption and the power system is relatively weak. Researching advanced information and data processing technologies to integrate and dispatch customer-side resources, and achieving two-way interaction between electricity and information, is of great significance for improving the efficiency and performance of the power system.

[0003] Researching energy consumption measurement methods for customer-side loads, effectively improving customer-side electricity efficiency, and implementing effective customer-side load management are considered important ways to solve the current energy shortage problem. Customer-side load management can guide electricity users to improve end-user electricity efficiency, optimize electricity consumption patterns, and scientifically allocate and dispatch power resources. This reduces electricity consumption and demand without affecting users' normal electricity needs, thereby minimizing peak-valley electricity consumption differences, reducing power transmission fluctuations on the distribution side, and improving the reliability and economy of the power grid.

[0004] Currently, there are two main methods for measuring customer-side load energy consumption: invasive load monitoring and non-invasive load monitoring. Invasive load monitoring is simple in principle, but it suffers from drawbacks such as high cost and the need for in-home sensor installation, leading to poor privacy. Non-invasive load monitoring, on the other hand, has been widely studied in recent years due to its wide user coverage and good data integrity. The key aspects of non-invasive load monitoring technology are load identification and load decomposition. With the continuous development of algorithms such as neural networks, there has been much research in load identification, and most methods have achieved high identification rates. However, research on load decomposition is relatively limited, and it does not achieve good results in load identification under conditions of multiple loads operating in parallel.

[0005] Therefore, it is necessary to study a load identification method that integrates multivariate information similarity analysis. Summary of the Invention

[0006] This invention proposes a load identification method and system based on comprehensive multi-source information similarity analysis to solve the problem of how to accurately determine the load input or output type.

[0007] To address the aforementioned problems, according to one aspect of the present invention, a load identification method integrating multivariate information similarity analysis is provided, the method comprising:

[0008] Determine whether there is a load change based on the fundamental amplitude corresponding to the current data in the current data window and the previous data window;

[0009] When a load change is confirmed and the change type is load shedding, the current data for the next data window is obtained;

[0010] Based on the current data of the previous data window and the current data of the next data window, load decomposition is performed to determine the first load decomposition data;

[0011] Multivariate information analysis is performed on the first load decomposition data and the second load decomposition data associated with the already put load to determine the first multivariate feature information vector corresponding to the first load decomposition data and the second multivariate feature information vector corresponding to each second load decomposition data.

[0012] The third load decomposition data corresponding to the cut-off load is determined based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector.

[0013] Based on the third load breakdown data, load matching is performed based on the correlation between the already put load and the second load breakdown data to determine the load to be cut off.

[0014] Preferably, determining whether there is a load change based on the fundamental amplitude corresponding to the current data of the current window and the previous data window includes:

[0015] If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out.

[0016] If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input.

[0017] If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

[0018] Preferably, the method further includes:

[0019] When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window;

[0020] Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point;

[0021] Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data;

[0022] Based on the second load decomposition data, load identification is performed to determine the type of input load and establish the correlation between the type of input load and the second load decomposition data.

[0023] Preferably, the method determines the multivariate feature information vector using the following methods:

[0024] For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the given load decomposition data is constructed.

[0025] Preferably, determining the third load decomposition data corresponding to the cut-off load based on the first multivariate feature information vector and the second multivariate feature information vector includes:

[0026] The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method:

[0027]

[0028] Where, d i Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between; x j Let y be the j-th element in q; ij For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector;

[0029] The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

[0030] According to another aspect of the present invention, a load identification system based on comprehensive multi-source information similarity analysis is provided, the system comprising:

[0031] The load change determination unit is used to determine whether there is a load change based on the fundamental amplitude value corresponding to the current data of the current window and the previous data window;

[0032] The data acquisition unit is used to acquire the current data of the next data window when it is determined that there is a load change and the change type is load shedding;

[0033] The first load decomposition data determination unit is used to perform load decomposition based on the current data of the previous data window and the current data of the next data window, and determine the first load decomposition data.

[0034] The multi-dimensional information analysis unit is used to perform multi-dimensional information analysis on the first load decomposition data and the second load decomposition data associated with the already put load, and to determine the first multi-dimensional feature information vector corresponding to the first load decomposition data and the second multi-dimensional feature information vector corresponding to each second load decomposition data.

[0035] The third load decomposition data determination unit is used to determine the third load decomposition data corresponding to the cut-off load based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector.

[0036] The load shedding determination unit is used to determine the load to be shedding based on the third load decomposition data and the correlation between the already-input load and the second load decomposition data.

[0037] Preferably, the load change determination unit determines whether a load change exists based on the fundamental amplitude corresponding to the current data of the current window and the previous data window, including:

[0038] If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out.

[0039] If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input.

[0040] If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

[0041] Preferably, the system further includes: an association establishment unit, used for:

[0042] When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window;

[0043] Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point;

[0044] Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data;

[0045] Based on the second load decomposition data, load identification is performed to determine the type of input load and establish the correlation between the type of input load and the second load decomposition data.

[0046] Preferably, the multivariate information analysis unit determines the multivariate feature information vector using the following method:

[0047] For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the given load decomposition data is constructed.

[0048] Preferably, the third load decomposition data determining unit determines the third load decomposition data corresponding to the cut-off load based on the first multivariate feature information vector and the second multivariate feature information vector, including:

[0049] The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method:

[0050]

[0051] Where, d i Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between; x j Let y be the j-th element in q; ij For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector;

[0052] The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

[0053] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the steps of a load identification method based on comprehensive multivariate information similarity analysis.

[0054] According to another aspect of the present invention, the present invention provides an electronic device, comprising:

[0055] The aforementioned computer-readable storage medium; and

[0056] One or more processors for executing a program in the computer-readable storage medium.

[0057] This invention provides a load identification method and system based on comprehensive multivariate information similarity analysis, comprising: determining whether a load change exists based on the fundamental amplitude corresponding to the current data of the current window and the previous data window; when a load change is determined to exist and the change type is load shedding, acquiring the current data of the next data window; performing load decomposition based on the current data of the previous data window and the current data of the next data window to determine first load decomposition data; performing multivariate information analysis on the first load decomposition data and the second load decomposition data associated with the already engaged load to determine a first multivariate feature information vector corresponding to the first load decomposition data and a second multivariate feature information vector corresponding to each second load decomposition data; determining third load decomposition data corresponding to the shedding load based on the first multivariate feature information vector and the second multivariate feature information vector; and performing load matching based on the correlation between the already engaged load and the second load decomposition data according to the third load decomposition data to determine the shedding load. The advantages of this invention are: (1) It can accurately realize load decomposition under the condition of multiple loads running in parallel, providing technical support for the application of non-intrusive load monitoring technology on the customer side and the establishment of equipment-level load monitoring system; (2) By analyzing the two data windows before and after the load switching time, load decomposition is performed, which solves the problem of the previous method of identifying multiple loads running in parallel as a new load, resulting in an excessively large training sample. The proposed method has certain engineering significance; (3) The load data is analyzed using fast Fourier transform, and load decomposition is achieved using harmonic analysis and reconstruction, which solves the problem of the previous method of directly analyzing the waveform in the data window. (4) The data window signals are analyzed from the time domain, frequency domain, and time-frequency domain respectively. The effective value, fundamental wave, third harmonic, fifth harmonic, seventh harmonic amplitude and wavelet packet energy entropy are calculated to construct a comprehensive multi-dimensional information feature vector. Compared with the previous method, which only analyzes data from the power threshold, it is more conducive to accurately judging load cut-out events. (5) The Euclidean distance is used to perform similarity analysis on the constructed comprehensive multi-dimensional information feature vector. The corresponding load reference data can be accurately judged and deleted when the load is cut off, so as to realize the update of the load reference dataset. Attached Figure Description

[0058] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0059] Figure 1 A flowchart of a load identification method 100 based on comprehensive multivariate information similarity analysis according to an embodiment of the present invention;

[0060] Figure 2 This is a flowchart illustrating load identification according to an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram illustrating load identification according to an embodiment of the present invention.

[0062] Figure 4 This is a schematic diagram of the electrical switching process according to an embodiment of the present invention;

[0063] Figure 5 This is a schematic diagram of the current waveforms before and after load connection according to an embodiment of the present invention;

[0064] Figure 6 This is a schematic diagram of the load identification system 600 based on comprehensive multi-source information similarity analysis according to an embodiment of the present invention. Detailed Implementation

[0065] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0066] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0067] Figure 1 This is a flowchart of a load identification method 100 based on comprehensive multi-source information similarity analysis according to an embodiment of the present invention. Figure 1As shown, the load identification method based on comprehensive multi-source information similarity analysis provided by the embodiments of the present invention (1) can accurately realize load decomposition under the condition of multiple loads running in parallel, providing technical support for the application of non-intrusive load monitoring technology on the customer side and the establishment of equipment-level load monitoring system; (2) by analyzing two data windows before and after the load switching time to decompose the load, it solves the problem of previous methods identifying multiple loads running in parallel as a new load, resulting in an excessively large training sample. The proposal of this method has certain engineering significance; (3) by using fast Fourier transform to analyze the load data and using harmonic analysis and reconstruction to realize load decomposition, it solves the problem of previous methods. The method directly subtracts the waveforms in the data window, but in practical applications, it is difficult to capture the phase difference between the waveforms in the two data windows; (4) The data window signals are analyzed from the time domain, frequency domain, and time-frequency domain respectively, and the effective value, fundamental wave, third harmonic, fifth harmonic, seventh harmonic amplitude and wavelet packet energy entropy are calculated to construct a comprehensive multi-dimensional information feature vector. Compared with the limitations of the previous method which only analyzes data from the power threshold alone, it is conducive to accurately judging the load cut-off event; (5) The Euclidean distance is used to perform similarity analysis on the constructed comprehensive multi-dimensional information feature vector, which can accurately judge and delete the corresponding load reference data when the load is cut off, and realize the update of the load reference dataset. The load identification method 100 of comprehensive multi-dimensional information similarity analysis provided by the embodiment of the present invention starts from step 101. In step 101, the fundamental wave amplitude corresponding to the current data of the current window and the previous data window is used to determine whether there is a load change.

[0068] Preferably, determining whether there is a load change based on the fundamental amplitude corresponding to the current data of the current window and the previous data window includes:

[0069] If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out.

[0070] If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input.

[0071] If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

[0072] Combination Figure 2As shown, in this invention, the data window length L is first set according to the sampling frequency, and current data D0 of one data window length is obtained according to the data window length L. After obtaining the data window D0, the data within the data window is subjected to spectrum analysis using Fast Fourier Transform (FFT) to obtain the fundamental amplitude FFT_D0 after decomposing the data window D0. Simultaneously, the current data D1 of the next data window is obtained, and the fundamental amplitude FFT_D1 after spectrum analysis of the corresponding data window is obtained. Then, FFT_D0 and FFT_D1 are compared to determine the specific load switching.

[0073] Specifically, if the absolute value of the difference between FFT_D0 and FFT_D1 is less than or equal to the preset threshold ε, it is determined that there is no load input or cut-off, and the data window is slid, replacing the data of data window D0 with the data of data window D1, and the subsequent calculation continues.

[0074] If the absolute value of the difference between FFT_D0 and FFT_D1 is greater than a preset threshold ε, then a load change is determined, indicating load switching or connection. The current value of the next data window D2 is acquired before determining whether a load has been switched on or off. Then, a threshold comparison is performed between FFT_D0 and FFT_D1 to clearly define the load switching event.

[0075] Specifically, if the difference between FFT_D1 and FFT_D0 is greater than the preset threshold ε, the load change type is determined to be load input; if the difference between FFT_D0 and FFT_D1 is greater than the preset threshold ε, the load change type is determined to be load cut-off.

[0076] In step 102, when it is determined that there is a load change and the change type is load cut-off, the current data of the next data window is obtained.

[0077] In step 103, load decomposition is performed based on the current data of the previous data window and the current data of the next data window to determine the first load decomposition data.

[0078] Combination Figure 2 As shown, in this invention, when a load cut-off is determined, the current data D2 of the next data window is obtained, and then the data D0 and D2 are analyzed. Considering that the amplitude of the current signal above the 8th harmonic is almost zero in actual situations, a fast Fourier transform is used to perform spectral analysis on the current data of the two data windows to obtain the amplitude from the fundamental wave to the 10th harmonic. Through linear calculation and reconstruction of 11 quantities, the load decomposition of the data is realized, and the first load decomposition data obtained from the load decomposition is saved as Q. In addition, after completing the load decomposition of the load cut-off signal, the sliding data window is used to assign values ​​from data D2 to D0.

[0079] In step 104, multivariate information analysis is performed on the first load decomposition data and the second load decomposition data associated with the already put loads, respectively, to determine the first multivariate feature information vector corresponding to the first load decomposition data and the second multivariate feature information vector corresponding to each second load decomposition data.

[0080] Preferably, the method determines the multivariate feature information vector using the following methods:

[0081] For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the given load decomposition data is constructed.

[0082] In step 105, the third load decomposition data corresponding to the cut-off load is determined based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector.

[0083] Preferably, determining the third load decomposition data corresponding to the cut-off load based on the first multivariate feature information vector and the second multivariate feature information vector includes:

[0084] The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method:

[0085]

[0086] Where, d i Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between; x j Let y be the j-th element in q; ij For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector;

[0087] The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

[0088] In step 106, load matching is performed based on the correlation between the already-input load and the second load decomposition data, according to the third load decomposition data, to determine the load to be cut off.

[0089] In this invention, when a load is applied, load decomposition is also performed to determine the second load decomposition data T. iThe system identifies loads based on load breakdown data and establishes a correlation between the identified loads and their corresponding second load breakdown data, saving this information for identifying loads that will be switched out when loads are switched out. Specifically, when loads are added, the second load breakdown data obtained from the load breakdown is saved as T. i Where i is related to the number of switching operations, with an initial value of 0, and i = i + 1 when the load is applied. After the load decomposition of the load application signal is completed, the sliding data window is used to assign values ​​from data D2 to D0.

[0090] Combination Figure 2 and Figure 3 As shown, in this invention, when a load cut-off occurs, the first load decomposition data Q and each stored second load decomposition data T are respectively processed. i Multivariate information analysis is performed, using time-domain feature calculation, fast Fourier transform, and wavelet packet transform to calculate the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes, as well as the wavelet packet energy entropy. A first multivariate feature information vector q and a second multivariate feature information vector t, combining time-domain, frequency-domain, and time-frequency-domain analysis, are then constructed. i After constructing a comprehensive multivariate feature information vector, the feature vector q and feature vector t are calculated. i The Euclidean distance d between them i Using Euclidean distance to measure data Q and data T i Similarity analysis, Euclidean distance d i The calculation formula is as follows:

[0091]

[0092] Where, d i Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between; x j Let y be the j-th element in q; ij For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector.

[0093] Then, based on the length of the Euclidean distance, the corresponding data T i Sort the data and select the load decomposition data T corresponding to the second multivariate feature information vector with the smallest Euclidean distance. k This is the third load decomposition data corresponding to the cut-off load. After determining the load decomposition data corresponding to the cut-off load, the data T with the smallest Euclidean distance is removed. k Simultaneously, due to the load shedding achieving i = i-1, data T is completed. iThe update is then performed. Finally, based on the third load breakdown data, load matching is conducted according to the correlation between the already engaged load and the second load breakdown data to determine the loads to be cut off.

[0094] Preferably, the method further includes:

[0095] When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window;

[0096] Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point;

[0097] Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data;

[0098] Based on the second load decomposition data, load identification is performed to determine the type of input load and establish the correlation between the type of input load and the second load decomposition data.

[0099] In this invention, when a load is applied, the current data D2 of the next data window is acquired, and then the data D0 and D2 are analyzed. Considering that the amplitude of the 8th and higher harmonics of the current signal is almost zero in actual situations, this patent uses Fast Fourier Transform to perform spectral analysis on the two data windows to obtain the amplitude of the fundamental to the 10th harmonic. Through linear calculation and reconstruction of 11 quantities, the load is decomposed to obtain the second load decomposed data, and the data obtained from the load decomposed data is saved as T. i Where i is related to the number of load switching operations, with an initial value of 0, and i = i + 1 when the load is applied. Finally, load identification is performed based on the second load decomposition data to determine the type of applied load, and a correlation is established between the type of applied load and the second load decomposition data.

[0100] In addition, after the load decomposition of the load input signal is completed, a sliding data window is needed to assign values ​​from data D2 to D0 in order to make judgments for the next cycle.

[0101] The method of this invention can be applied to the research and development and configuration of load identification devices for customer-side power equipment in low-voltage systems, and can provide technical support for load identification in complex power consumption scenarios where multiple loads operate in parallel.

[0102] The following specific examples illustrate the embodiments of the present invention.

[0103] In the embodiments of the present invention, the method is actually verified using two electrical appliances: a small sun heater and an electric drill.

[0104] In a specific example verification, a sampling frequency of 100kHz was used to collect voltage and current signals, and the collected data was divided into data windows with a length L = 6000. To verify that the decomposed waveform still represents the current waveform of the appliance and has identifiable significance, the current waveforms collected under the individual operating conditions of four appliances—a small electric heater, a kettle, an electric drill, and a space heater—were used as input to train the network model before starting. This training was then used to identify the load on the decomposed waveform, thus verifying the effectiveness of the proposed method.

[0105] To verify that this method can achieve load decomposition, a design was created as follows: Figure 4 The electrical appliance switching process is shown below. First, the line is kept unloaded. At time t1, a small heater is connected and runs continuously until time t2. After time t2, a drill is connected, and both appliances run in parallel until time t3. Then, the small heater is disconnected, leaving only the drill running alone until time t4. Finally, the drill is disconnected, and the line returns to an unloaded state. The load decomposition process will be described in detail below.

[0106] First, two current data points of length L are read from the real-time sampling data, and the data read from the first data window is saved as standard data under no-load conditions. Next, a spectrum analysis is performed on the data from the two data windows. If the absolute value of the difference between the fundamental amplitudes of the data is less than the threshold ε, it proves that there is no load switching. Then, the data windows are slid to update the data from the two data windows, and the load switching judgment is made at the next time step.

[0107] Until time t1, after data window spectrum analysis, a load connection event is identified. The next data window is read to obtain the steady-state operating data after load A is connected. The newly read data window data is analyzed with the data window data before load connection, and the fundamental to 10th harmonic data of the two data windows are extracted. The corresponding frequency band data is linearly processed and reconstructed. The processed information is saved as the standard data of load A for subsequent analysis and judgment when the load is disconnected. After completing the first load connection judgment process, the steady-state data window data after load connection, the standard data of no-load state, and the standard data of load A are subjected to harmonic extraction and decomposition processing to obtain the decomposed data at that time. Then, the decomposed data is sent to the previously trained GRU network for load identification. The algorithm identifies load A as "small sun" and establishes the correlation between "small sun" and the corresponding decomposed data. After completing the above steps, the data window is slid to judge the next connection event.

[0108] At time t2, after data window spectrum analysis, it was determined that load B had been connected. Data processing was performed as described in the previous steps. After acquiring the standard data for load B, the steady-state data window after the load was connected was selected and analyzed along with the standard data for each load to complete the waveform decomposition of the multiple loads. The waveform comparison before and after the load connection was as follows: Figure 5 As shown. Next, the decomposed signal is sent to the identification program for identification, confirming that an electric drill is connected. Then, the sliding data window is used for subsequent event detection and analysis.

[0109] At time t3, after data window spectrum analysis, it is determined that a load has been cut off. The steady-state data after the load cut off is then read and subjected to harmonic extraction and decomposition processing with the data window data before the cut-off, yielding standard data for the cut-off load. Next, this standard data is analyzed along with the standard data for loads A and B, extracting the effective value, harmonic amplitude, and wavelet packet energy entropy from the data window data to construct corresponding multivariate information feature vectors. Then, the Euclidean distance between the multivariate information feature vector of the cut-off load and the corresponding feature vectors of the standard data for loads A and B is calculated. The load standard data with the closest Euclidean distance is identified as the cut-off load, and its corresponding standard data is removed, completing the update of the load standard database. After updating the standard database, the steady-state data after the cut-off event and the load standard data are subjected to harmonic analysis and reconstruction processing to obtain the current decomposition data at the current time. The determination of the cut-off load is then based on the correlation between the load and the decomposition data. After completing the above steps, the sliding data window continues for subsequent analysis.

[0110] At time t4, after data window spectrum analysis, it is determined that a load has been cut off. As described in the load cut-off determination steps at time t3, the cut-off event of the electric drill is determined. Finally, experiments were conducted by setting up separate and parallel operation modes for two types of loads, thus verifying the effectiveness of the method of this invention. Therefore, the method of this invention can achieve load identification under conditions of parallel operation of multiple electrical appliances.

[0111] Figure 6 This is a schematic diagram of the load identification system 700 based on comprehensive multi-source information similarity analysis according to an embodiment of the present invention. Figure 6 As shown, the load identification system 600 based on comprehensive multi-dimensional information similarity analysis provided in this embodiment of the invention includes: a load change determination unit 601, a data acquisition unit 602, a first load decomposition data determination unit 603, a multi-dimensional information analysis unit 604, a third load decomposition data determination unit 605, and a load cut-off determination unit 606.

[0112] Preferably, the load change determination unit 601 is used to determine whether there is a load change based on the fundamental amplitude value corresponding to the current data of the current window and the previous data window.

[0113] Preferably, the load change determination unit 601 determines whether a load change exists based on the fundamental amplitude corresponding to the current data of the current window and the previous data window, including:

[0114] If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out.

[0115] If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input.

[0116] If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

[0117] Preferably, the data acquisition unit 602 is used to acquire the current data of the next data window when it is determined that there is a load change and the change type is load shedding.

[0118] Preferably, the first load decomposition data determination unit 603 is used to perform load decomposition based on the current data of the previous data window and the current data of the next data window to determine the first load decomposition data.

[0119] Preferably, the multivariate information analysis unit 604 is used to perform multivariate information analysis on the first load decomposition data and the second load decomposition data associated with the already put load, respectively, to determine the first multivariate feature information vector corresponding to the first load decomposition data and the second multivariate feature information vector corresponding to each second load decomposition data.

[0120] Preferably, the multivariate information analysis unit 604 determines the multivariate feature information vector using the following method:

[0121] For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the given load decomposition data is constructed.

[0122] Preferably, the system further includes: an association establishment unit, used for:

[0123] When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window;

[0124] Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point;

[0125] Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data;

[0126] Based on the second load decomposition data, load identification is performed to determine the type of input load and establish the correlation between the type of input load and the second load decomposition data.

[0127] Preferably, the third load decomposition data determination unit 605 is used to determine the third load decomposition data corresponding to the cut-off load based on the first multivariate feature information vector and the second multivariate feature information vector.

[0128] Preferably, the load cut-out determination unit 606 is used to determine the load cut-out based on the third load decomposition data and the correlation between the already-input load and the second load decomposition data.

[0129] Preferably, the third load decomposition data determination unit 605 determines the third load decomposition data corresponding to the cut-off load based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector, including:

[0130] The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method:

[0131]

[0132] Where, d i Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between; x j Let y be the j-th element in q; ij For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector;

[0133] The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

[0134] The load identification system 600 based on comprehensive multivariate information similarity analysis in this embodiment corresponds to the load identification method 100 based on comprehensive multivariate information similarity analysis in another embodiment of this invention, and will not be described again here.

[0135] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the steps of a load identification method based on comprehensive multivariate information similarity analysis.

[0136] According to another aspect of the present invention, the present invention provides an electronic device, comprising:

[0137] The aforementioned computer-readable storage medium; and

[0138] One or more processors for executing a program in the computer-readable storage medium.

[0139] The invention has been described with reference to a few embodiments. However, as will be known to those skilled in the art, and as defined in the appended claims, other embodiments besides those disclosed above fall equivalently within the scope of the invention.

[0140] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the art, unless otherwise expressly defined herein. All references to “a / the / the [device, component, etc.]” ​​are openly interpreted as at least one instance of said device, component, etc., unless otherwise expressly stated. The steps of any method disclosed herein need not be performed in the exact order disclosed unless explicitly stated otherwise.

[0141] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0143] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0144] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A load identification method based on comprehensive multi-source information similarity analysis, characterized in that, The method includes: Determine whether there is a load change based on the fundamental amplitude corresponding to the current data in the current data window and the previous data window; When a load change is confirmed and the change type is load shedding, the current data for the next data window is obtained; Based on the current data of the previous data window and the current data of the next data window, load decomposition is performed to determine the first load decomposition data; Multivariate information analysis is performed on the first load decomposition data and the second load decomposition data associated with the already put load to determine the first multivariate feature information vector corresponding to the first load decomposition data and the second multivariate feature information vector corresponding to each second load decomposition data. The third load decomposition data corresponding to the cut-off load is determined based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector. Based on the third load breakdown data, load matching is performed based on the correlation between the already engaged load and the second load breakdown data to determine the load to be cut off. The method further includes: When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window; Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point; Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data; Based on the second load decomposition data, load identification is performed to determine the type of input load and to establish the correlation between the type of input load and the second load decomposition data; The method determines the multivariate feature information vector in the following manner: For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the load decomposition data is constructed.

2. The method according to claim 1, characterized in that, The determination of whether there is a load change based on the fundamental amplitude corresponding to the current data of the current window and the previous data window includes: If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out. If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input. If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

3. The method according to claim 1, characterized in that, The step of determining the third load decomposition data corresponding to the cut-off load based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector includes: The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method: ， in, Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between them; Let be the j-th element in q; For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector; The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

4. A load identification system based on comprehensive multi-source information similarity analysis, characterized in that, The system includes: The load change determination unit is used to determine whether there is a load change based on the fundamental amplitude corresponding to the current data of the current window and the previous data window; The data acquisition unit is used to acquire the current data of the next data window when it is determined that there is a load change and the change type is load shedding; The first load decomposition data determination unit is used to perform load decomposition based on the current data of the previous data window and the current data of the next data window, and determine the first load decomposition data. The multi-dimensional information analysis unit is used to perform multi-dimensional information analysis on the first load decomposition data and the second load decomposition data associated with the already put load, and to determine the first multi-dimensional feature information vector corresponding to the first load decomposition data and the second multi-dimensional feature information vector corresponding to each second load decomposition data. The third load decomposition data determination unit is used to determine the third load decomposition data corresponding to the cut-off load based on the first multi-dimensional feature information vector and the second multi-dimensional feature information vector. The load cut-out determination unit is used to determine the load to be cut out based on the third load decomposition data and the correlation between the already-in load and the second load decomposition data by performing load matching. The system further includes: an association establishment unit, used for: When a load change is confirmed, and the change type is load input, obtain the current data D2 of the next data window; Perform spectrum analysis on the current data of the previous data window and the current data of the next data window respectively to obtain the amplitude of the fundamental wave to the preset harmonic for each current data point; Linear calculation and reconstruction are performed based on the amplitude of the fundamental wave to the preset harmonic for each current data to obtain the second load decomposition data; Based on the second load decomposition data, load identification is performed to determine the type of input load and to establish the correlation between the type of input load and the second load decomposition data; The multi-dimensional information analysis unit determines the multi-dimensional feature information vector using the following methods: For any load decomposition data, the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy are calculated using time-domain features, fast Fourier transform, and wavelet packet transform. Based on the effective value, fundamental frequency, third harmonic, fifth harmonic, and seventh harmonic amplitudes and wavelet packet energy entropy, a multivariate feature information vector corresponding to the load decomposition data is constructed.

5. The system according to claim 4, characterized in that, The load change determination unit determines whether a load change exists based on the fundamental amplitude corresponding to the current data in the current window and the previous data window, including: If the difference between the fundamental amplitude FFT_D0 corresponding to the current data D0 of the previous data window and the fundamental amplitude FFT_D1 corresponding to the current data D1 of the current window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load cut-out. If the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is greater than a preset threshold, then it is determined that there is a load change, and the change type is load input. If the absolute value of the difference between the fundamental amplitude FFT_D1 corresponding to the current data D1 in the current data window and the fundamental amplitude FFT_D0 corresponding to the current data D0 in the previous data window is less than or equal to a preset threshold, then it is determined that there is no load change.

6. The system according to claim 4, characterized in that, The third load decomposition data determination unit determines the third load decomposition data corresponding to the cut-off load based on the first multivariate feature information vector and the second multivariate feature information vector, including: The Euclidean distance between the first multivariate feature information vector and each of the second multivariate feature information vectors is calculated using the following method: ， in, Let q be the first multivariate feature information vector and T be the i-th second multivariate feature information vector. i Euclidean distance between them; Let be the j-th element in q; For T i The j-th element in the feature information vector; n is the number of elements in the feature information vector; The load decomposition data corresponding to the second multivariate feature information vector with the smallest Euclidean distance is selected as the third load decomposition data corresponding to the cut-off load.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-3.

8. An electronic device, characterized in that, include: The computer-readable storage medium as described in claim 7; as well as One or more processors for executing a program in the computer-readable storage medium.

Images