Non-intrusive load identification method and system for homogeneous industrial load

Abstract

The invention discloses a non-intrusive load identification method and a non-intrusive load identification system for a homogeneous industrial load. The non-intrusive load identification method comprises the following steps of performing multi-dimensional segmented aggregation approximation processing on current data # imgabs0 # in high-frequency data of the industrial load after high-frequency centralization processing; constructing a distance similarity matrix by using the data subjected to the multi-dimensional segmented aggregation approximation processing; converting the distance similarity matrix into a weighted recursive matrix through normalization and threshold control; constructing a discretized V-I track based on high-frequency current and voltage data in the directly obtained high-frequency data of the industrial load; based on the weighted recursive matrix, performing weighted mapping on the discretized industrial load V-I track; obtaining a recursive V-I trajectory feature map after mapping by using different types of homogeneous industrial load data, and training to obtain a load classifier; and classifying the to-be-classified industrial loads by using the load classifier. According to the method, the industrial loads with relatively high homogenization can be effectively identified, and the accuracy is relatively high.

Description

Technical Field

[0001] This invention belongs to the field of load monitoring and analysis in industrial power systems, and relates to a non-intrusive load identification method and system for homogeneous industrial loads. Background Technology

[0002] With the continuous development of digital grid technology, energy visualization, reduction of overall energy consumption, and peak shaving and valley filling are urgent needs for building new power systems, and online monitoring of power load is key to achieving these goals. There are two main methods for online power load monitoring: Intrusive Load Monitoring (ILM) and Non-Intrusive Load Monitoring (NILM). NILM has a simple hardware structure, is easy to install and maintain, and is more suitable for the actual needs of smart grid construction.

[0003] Current non-intrusive load identification methods mainly focus on the residential side, with limited research on industrial loads due to difficulties in data collection. However, the stability, controllability, and large proportion of industrial loads are crucial for building new power systems. Existing research primarily relies on low-frequency data combined with event analysis to determine load types. This approach suffers from drawbacks such as excessively long intervals between load events, resulting in lost load information, and difficulty in extracting effective event occurrence and termination markers due to the high frequency of load events, leading to identification errors. Furthermore, low-frequency data exhibits time lag effects, making it difficult to handle the refined scheduling of today's complex power systems. Industrial loads are highly homogeneous, with a large number of motor-type loads. Research based on high-frequency data has limited ability to differentiate between these homogeneous loads, making misjudgments likely.

[0004] In existing research, feature representation based solely on voltage-current (VI) trajectories is insufficient to fully capture the time-series characteristics of voltage and current, especially when dealing with loads with similar VI trajectories, exhibiting significant limitations. Furthermore, although some studies have employed algorithms specifically designed for dynamic feature extraction of time series data, such as weighted recursive graphs, Markov transition fields, and Gram angle fields, and proposed load feature representation and identification methods that integrate these approaches, these techniques may suffer from excessively high information dimensionality during feature extraction, leading to the coverage or partial loss of features from homogeneous loads. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a non-intrusive identification and classification method for industrial loads based on high-frequency current and voltage data. Building upon traditional VI trajectory features, it further introduces time-series feature extraction and avoids the overwriting and loss of redundant feature information by capturing key features through high-dimensional spatial mapping. Specifically, it maps these features to the RGB color space to generate new feature maps, thereby achieving more accurate load identification. This method is particularly suitable for the precise monitoring and classification of highly homogeneous industrial loads in new power systems.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a non-intrusive load identification method for homogeneous industrial loads, comprising the following steps: Current data in high-frequency data of industrial load after high-frequency centralization processing Perform multi-dimensional segmented aggregation approximation processing; A distance similarity matrix is ​​constructed using the data after the multi-dimensional segmented aggregation approximation process. The similarity of load characteristics is quantified by calculating the distance between data points; By normalizing and thresholding, the distance similarity matrix is... Convert to a weighted recursive matrix Extract key information about load characteristics; Discretized VI trajectories are constructed based on high-frequency current and voltage data from directly acquired high-frequency industrial load data. Based on the weighted recursive matrix The discrete industrial load VI trajectory is weighted and mapped to form a recursive VI trajectory feature map after mapping. A load classifier is obtained by training recursive VI trajectory feature maps after mapping from different categories of homogeneous industrial load data. Based on the industrial load to be classified, the load classifier is used to classify the industrial load to be classified.

[0007] Furthermore, the high-frequency data of the industrial load includes voltage data and current data, and the storage matrix is ​​as follows: and For the matrix as and Centralized processing includes: Calculate the minimum values ​​of current and voltage signals. , and maximum value , ; Step size for calculating data : ,in for , ; Based on step size Adjust the current and voltage signals to make their data distribution more uniform.

[0008] Furthermore, regarding the current data When performing multidimensional piecewise aggregation approximation, The mean of each signal segment is calculated as follows:

[0009] in , For segment index, The length of the original signal. Set the target embedding size.

[0010] Furthermore, construct a distance similarity matrix. This includes constructing a distance metric matrix between each time series point, where the matrix metric uses Euclidean distance; the distance between any two points is defined as follows: ,in, This is the power parameter used in distance calculation.

[0011] Furthermore, through normalization and threshold control, the distance similarity matrix is... Convert to a weighted recursive matrix Key information for extracting load characteristics includes: Calculate the normalized distance matrix : ; Limit the distance value to a threshold. Within: ; Distance similarity matrix Transform into a weighted recursive graph ,parameter and The selection satisfies ,in, .

[0012] Furthermore, based on the weighted recursive matrix The discrete industrial load VI trajectory is weighted and mapped to form a recursive VI trajectory feature map. The main features are mapped from low-dimensional space to high-dimensional space through high-dimensional mapping technology, feature decomposition is performed, the feature with the greatest feature difference in high-dimensional space is captured and projected in low dimension, the projected data is normalized, and color mapping is used to realize the visualization of the difference features, so as to obtain an improved VI trajectory map that can distinguish homogeneous industrial loads.

[0013] Furthermore, by generating five load feature datasets, which are proportionally divided into training, testing, and validation sets, each feature dataset contains several feature vectors, which represent the feature information of different loads; and based on the classic neural network parametric residual network, the load classifier is trained using the training, testing, and validation sets.

[0014] Based on the concept of the method, the present invention also provides a non-intrusive load identification system for homogeneous industrial loads, comprising a data processing module, a similarity calculation module, a feature extraction module, a VI trajectory acquisition module, a feature map construction module, a classifier training module, and a classification module; The data processing module is used to process the current data in the high-frequency data of industrial loads after high-frequency centralization. Perform multi-dimensional segmented aggregation approximation processing; The similarity calculation module uses the data after multi-dimensional segmentation and aggregation approximation to construct a distance similarity matrix. The similarity of load characteristics is quantified by calculating the distance between data points; The feature extraction module uses normalization and threshold control to transform the distance similarity matrix. Convert to a weighted recursive matrix Extract key information about load characteristics; The VI trajectory acquisition module is used to construct discretized VI trajectories based on high-frequency current and voltage data from directly acquired high-frequency industrial load data; The feature map construction module is used to construct the feature map based on the weighted recursive matrix. The discrete industrial load VI trajectory is weighted and mapped to form a recursive VI trajectory feature map after mapping. The classifier training module uses the recursive VI trajectory feature map after mapping the homogeneous industrial load data of different categories to train the load classifier. The classification module classifies the industrial load to be classified using the load classifier.

[0015] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the non-intrusive load identification method for homogeneous industrial loads described in the present invention when executing the computer executable program.

[0016] Simultaneously, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, it can implement the non-intrusive load identification method for homogeneous industrial loads described in this invention.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: This invention reveals the periodicity and complexity of industrial load time series data while maintaining intuitiveness, enabling effective identification of industrial loads with high accuracy and enriching the application areas and scope of load monitoring. Building upon traditional VI trajectory features, it further introduces time series feature extraction, using a high-dimensional space mapping method to capture key features and avoid the overwriting and loss problems caused by redundant feature information. Specifically, through multi-dimensional PAA processing technology, this invention can significantly reduce data dimensionality and improve data processing efficiency while maintaining key feature information. By constructing a weighted recursive matrix, key feature information of the load is further extracted, enabling more accurate differentiation of highly homogeneous industrial loads. Based on the key information extracted from time series data using the weighted recursive matrix, this invention reprojects the key information into a high-dimensional space to decompose its dimensional features, capturing the dimensionality that maximizes feature differences. Utilizing color mapping of differences, an improved VI trajectory diagram is constructed to intuitively display the dynamic characteristics of the load. This visualization method helps to observe and analyze load characteristics more intuitively, assisting users in effective load management and monitoring in complex power systems; and maps the VI trajectory map to the RGB color space to generate new feature maps, thereby achieving more accurate load identification, which is particularly suitable for the accurate monitoring and classification of highly homogeneous industrial loads in new power systems. Attached Figure Description

[0018] The present invention will become apparent and readily understood from the following description of embodiments taken in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart of the classification of loads in homogeneous industries.

[0019] Figure 2 This is a comparison diagram of the original VI trajectories of five types of homogeneous industrial loads according to the present invention.

[0020] Figure 3 Comparison chart of the VI trajectory for the first type of industrial load improvement.

[0021] Figure 4 Comparison chart of the VI trajectory for the second type of industrial load improvement.

[0022] Figure 5 Comparison chart of the VI trajectory for the third type of industrial load improvement.

[0023] Figure 6 Comparison chart of the VI trajectory for the fourth type of industrial load improvement.

[0024] Figure 7 Comparison chart of the fifth category of industrial load improvement VI trajectory. Detailed Implementation

[0025] The embodiments of the present invention are described in detail below, wherein the same or similar reference numerals denote the same variables throughout.

[0026] This invention provides a non-intrusive load identification method for homogeneous industrial loads, comprising the following steps: Step 1: Collect high-frequency data of industrial load from the system, store it as a matrix, and then perform centralized processing.

[0027] In step 1, high-frequency current and voltage data are acquired from the industrial power system, periodically decomposed, and then stored in a matrix. and In the matrix, each row represents a period, and each column corresponds to a different time point. The acquired voltage and current data are centralized to ensure data balance in subsequent processing. The centralization steps are as follows: (1) Calculate the minimum values ​​of the current and voltage signals. , and maximum value , .

[0028] (2) Calculate the step size of the data : ,in for , .

[0029] (3) Based on step size Adjust the current and voltage signals to make their data distribution more uniform.

[0030] Step 2, based on the high-frequency centralized current data of the industrial load. Multidimensional segmented aggregation approximation (PAA) processing is performed; in this step, the centered current signal is... PAA processing is performed to compress the signal dimensionality, reduce the data volume, and retain the main characteristics of the load. Let the length of the original signal be... The target embedding size is The mean of each signal segment is calculated as follows:

[0031] in, , For segment indexes.

[0032] Step 3: Construct a distance similarity matrix using the PAA-based processed data. The similarity of load characteristics is quantified by calculating the distance between data points; In this step, the current signal is based on PAA processing. Calculate the distance between each pair of data points and construct a distance similarity matrix. The distance between any two points is defined as follows: ,in, This is the power parameter in distance calculation, used to control how the distance is measured. The final distance similarity matrix is ​​obtained. .

[0033] Step 4: Based on the normalization and threshold control described above, the distance similarity matrix is... Convert to a weighted recursive matrix To extract key information about load characteristics; In this step, the distance similarity matrix is ​​used. Construct a weighted recursive matrix By setting appropriate parameters and The matrix is ​​weighted to improve the discriminative power of load characteristics. The specific weighting steps are as follows: First, calculate the normalized distance matrix. : ; Then limit the distance value to a threshold. Within: .

[0034] The distance similarity matrix is ​​obtained using the method described above. Transform into a weighted recursive graph , where parameters and The selection must meet the following requirements. ,in, .

[0035] Step 5: Construct the corresponding VI trajectory from the collected high-frequency current and voltage data of the industrial load; In this step, the high-frequency current and voltage data of the industrial load collected are unprocessed data with a dimension of 1× .

[0036] Step 6: Based on the weighted recursive matrix The discretized industrial load VI trajectory is differentially mapped to form a new recursive VI trajectory feature map. In this step, the dimension of the discretized industrial load VI trajectory diagram is... The VI trajectory is composed of Composed of discrete points of high-frequency voltage and current, weighted recursive matrix The dimension is m×m, and the difference mapping is performed by capturing the most dissimilar feature in the high-dimensional space.

[0037] The mapping process is as follows:

[0038] Among them, high-dimensional mapping The mapping process: (1) To facilitate calculation, the mapping function will be... Defined as ; (2) After centralization Perform eigenvalue decomposition ; (3) Calculate the principal features in the feature space ,in ; (4) Obtain the final projection of the low-dimensional space. We obtain the low-dimensional main features after projection.

[0039] Normalize the projected low-dimensional main features:

[0040] Assign the normalized principal features at each level to the RGB channels:

[0041] The final result is an improved VI trajectory diagram that can distinguish homogeneous industrial loads.

[0042] Step 7: Use the feature maps obtained from the homogeneous industrial load data of each category to train a load classifier; In this step, five load feature datasets are generated through steps 1-6 to train the load classifier (refer to 2). These feature datasets are divided into training set, test set and validation set according to the proportion. Each feature dataset contains several feature vectors, which represent the feature information of different loads. The load classifier is trained through the training set and test set. To highlight the universality of the proposed method, a classic neural network parametric residual network is used.

[0043] Step 8: Based on the industrial load to be classified, classify the industrial load to be classified using the load classifier.

[0044] Figure 2 A comparative chart of voltage-current (VI) trajectory characteristics for five industrial loads (File 1 to File 5) is presented. It is evident that the voltage and current curves of these industrial loads at each moment exhibit high similarity, and their trends are almost identical. Therefore, it can be concluded that traditional VI trajectory plots have significant limitations in distinguishing highly homogeneous industrial loads.

[0045] Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 The application effects of the method of this invention are demonstrated. This method first extracts time-series features from voltage and current data using a distance similarity algorithm. Then, through high-dimensional spatial mapping, the three main extracted features are mapped to RGB color channels. In this process, the R channel corresponds to the most important feature, with 255 chromaticity values. The larger the chromaticity value, the greater the fluctuation of the feature, the greater the distance, and the lower the similarity. The G and B channels correspond to the second and third most important features, respectively, and the principle is the same as that of the R channel. Over time, the chromaticity changes in the graph intuitively reflect the dynamic differences of each load feature. The chromaticity changes amplify the feature fluctuations that cannot be expressed in the VI trajectory, effectively improving the clarity of feature expression and the accuracy of load identification.

[0046] This invention provides a non-invasive method for identifying and classifying industrial loads based on high-frequency current and voltage data. It aims to address the problem of low accuracy in identifying homogeneous industrial loads in existing technologies. Through multi-dimensional data processing, feature extraction and recursive graph construction, and a deep learning classification model, it achieves accurate identification and classification of complex industrial loads, offering the following advantages: (1) High-precision load feature extraction By employing multidimensional PAA processing technology, this invention can significantly reduce data dimensionality and improve data processing efficiency while preserving key feature information. Furthermore, by constructing a weighted recursive matrix, key feature information of the load is extracted, enabling more accurate differentiation of highly homogeneous industrial loads.

[0047] (2) Intuitive visualization of dynamic features This invention, building upon the extraction of key information from time-series data using a weighted recursive matrix, reprojects this key information into a high-dimensional space to decompose its features across various dimensions. It captures the dimension maximizing feature differences and utilizes color mapping to construct an improved VI trajectory diagram that visually displays the dynamic characteristics of the load. This visualization method facilitates more intuitive observation and analysis of load characteristics, assisting users in effective load management and monitoring within complex power systems.

[0048] (3) Precise classification of homogeneous industrial loads This invention, by designing and training a classical neural network residual network, can effectively address the challenge of identifying homogeneous industrial loads, significantly improve classification accuracy, and contribute to the realization of refined scheduling and energy visualization management of power systems.

[0049] In addition, using the same technical concept as the method described above, a non-intrusive load identification system for homogeneous industrial loads is provided, comprising a data processing module, a similarity calculation module, a feature extraction module, a VI trajectory acquisition module, a feature map construction module, a classifier training module, and a classification module; The data processing module is used to process the current data in the high-frequency data of industrial loads after high-frequency centralization. Perform multi-dimensional segmented aggregation approximation processing; The similarity calculation module uses the data after multi-dimensional segmentation and aggregation approximation to construct a distance similarity matrix. The similarity of load characteristics is quantified by calculating the distance between data points; The feature extraction module uses normalization and threshold control to transform the distance similarity matrix. Convert to a weighted recursive matrix Extract key information about load characteristics; The VI trajectory acquisition module is used to construct discretized VI trajectories based on high-frequency current and voltage data from directly acquired high-frequency industrial load data; The feature map construction module is used to construct the feature map based on the weighted recursive matrix. The discrete industrial load VI trajectory is weighted and mapped to form a recursive VI trajectory feature map after mapping. The classifier training module uses the recursive VI trajectory feature map after mapping the homogeneous industrial load data of different categories to train the load classifier. The classification module classifies the industrial load to be classified using the load classifier.

[0050] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the non-intrusive load identification method for homogeneous industrial loads described in the present invention when executing the computer executable program.

[0051] On the other hand, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the implementation of the non-intrusive load identification method for homogeneous industrial loads described in the present invention.

[0052] The computer device may be a laptop, a desktop computer, or a workstation.

[0053] The processor can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).

[0054] The memory described in this invention can be an internal storage unit of a laptop, desktop computer, or workstation, such as memory or hard disk; or it can be an external storage unit, such as a portable hard disk or flash memory card.

[0055] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0056] The above content is only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A non-intrusive load identification method for homogeneous industrial loads, characterized in that: The following steps are involved: Current data in high-frequency data of industrial load after high-frequency centralization processing Perform multi-dimensional segmented aggregation approximation processing; The distance similarity matrix is ​​constructed using the data processed by the multi-dimensional segmented aggregation approximation. , quantifies the similarity of load characteristics by calculating the distance between data points; Through normalization and threshold control, the distance similarity matrix Convert to a weighted recursive matrix , extract key information of load characteristics; Constructing discretized VI trajectory based on high-frequency current and voltage data directly obtained from high-frequency data of industrial loads; Based on the weighted recursive matrix , weighted mapping is performed on the discretized industrial load VI trajectory to form a recursive VI trajectory feature map after mapping; The load classifier is obtained by using the recursive VI trajectory feature map after mapping using homogeneous industrial load data of different categories for training. Based on the industrial load to be classified, the industrial load to be classified is classified by using the load classifier.

2. The non-intrusive load identification method for homogenized industrial loads according to claim 1 is characterized in that: The high-frequency data of the industrial load includes voltage data and current data, and the storage matrix is and ; For the matrix and Centralized processing, including: Calculate the minimum value of current and voltage signals , and maximum value , ; Calculate the step size of the data : ,in for , ; According to the step length Adjust the current and voltage signals to make their data distribution more uniform.

3. The non-intrusive load identification method for homogenized industrial loads according to claim 1 is characterized in that: Current data When performing multi-dimensional segmented aggregation approximation processing, The mean of each segment of the signal is calculated as follows: in , is the segment index, is the length of the original signal, is the target embedding size.

4. The non-intrusive load identification method for homogenized industrial loads according to claim 1 is characterized in that: Constructing distance similarity matrix It includes constructing a distance measurement matrix between each time series point, where the matrix measurement uses Euclidean distance; the distance between any two points is defined as follows: ,in, is the power parameter in the distance calculation.

5. The non-intrusive load identification method for homogenized industrial loads according to claim 1 is characterized in that: Through normalization and threshold control, the distance similarity matrix Convert to a weighted recursive matrix ,The key information of load characteristics extraction includes: Calculate the normalized distance matrix : ; Clamp distance values ​​to a threshold Within: ; The distance similarity matrix Convert to a weighted recursive graph ,parameter and The selection satisfies ,in, .

6. The non-intrusive load identification method for homogenized industrial loads according to claim 1 is characterized in that: Based on the weighted recursive matrix , the discretized industrial load VI trajectory is weighted mapped to form a recursive VI trajectory feature map after mapping, including: the main feature is mapped from low-dimensional space to high-dimensional space through high-dimensional mapping technology, feature decomposition is performed, the maximum feature of feature differentiation in high-dimensional space is captured for low-dimensional projection, the projection data is normalized, and color mapping is used to realize the visualization expression of difference features, so as to obtain an improved VI trajectory map that can distinguish homogeneous industrial loads.

7. The non-intrusive load identification method for homogenized industrial loads according to claim 1, characterized in that: By generating five load characteristic data sets, the characteristic data sets are divided into training set, test set and validation set in proportion. Each characteristic data set contains several feature vectors, and the feature vectors represent the characteristic information of different loads. Based on the classical neural network residual network, the load classifier is trained through the training set, test set and validation set.

8. A non-intrusive load identification system for homogeneous industrial loads, characterized in that: Data processing module, similarity calculation module, feature extraction module, VI trajectory acquisition module, feature map construction module, classifier training module and classification module; The data processing module is used to process the current data of the high-frequency data of industrial loads after high-frequency central processing Perform multi-dimensional segmented aggregation approximation processing; The similarity calculation module uses the data after the multi-dimensional segmented aggregation approximation processing to construct a distance similarity matrix , quantifies the similarity of load characteristics by calculating the distance between data points; The feature extraction module transforms the distance similarity matrix into Convert to a weighted recursive matrix , extract key information of load characteristics; The VI trajectory acquisition module is used to construct a discretized VI trajectory based on high-frequency current and voltage data in the high-frequency data of the industrial load directly acquired; The feature map construction module is used based on the weighted recursive matrix , weighted mapping is performed on the discretized industrial load VI trajectory to form a recursive VI trajectory feature map after mapping; The classifier training module uses the recursive VI trajectory feature map after mapping by using the homogeneous industrial load data of different categories to train the load classifier; The classification module classifies the industrial load to be classified based on the industrial load to be classified by using the load classifier.

9. A computer device, characterized in that: It includes a processor and a memory, the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and when the processor executes the computer executable program, it can implement the non-intrusive load identification method for homogeneous industrial loads as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored in a computer-readable storage medium. When the computer program is executed by a processor, the non-intrusive load identification method for homogeneous industrial loads as described in any one of claims 1 to 7 can be implemented.

Images