Power system load type intelligent identification method, system and equipment

Abstract

The invention relates to the field of power systems, in particular to a power system load type intelligent identification method, system and equipment. The method comprises the steps that clustering centers of a plurality of load type subgroups are determined respectively, and each clustering center represents a typical load type calibrated in advance; utilizing a dynamic time warping (DTW) algorithm to calculate a DTW distance between the to-be-identified load node and each clustering center; according to the DTW distance obtained through calculation, distributing the to-be-identified load node to a subgroup of the nearest clustering center, and determining the load type of the to-be-identified load node; and periodically scanning and identifying load type changes of the load units in the domain, and updating a load type identification result. The invention further comprises a power system load type intelligent identification system and electronic equipment, the identification accuracy of the multi-type industrial load composite bus is improved, and misjudgment and deviation are reduced. The bus load type can be dynamically identified and changed and can be updated in time.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to a method, system and equipment for intelligently identifying load types in power systems. Background Art

[0002] Intelligent identification of load types in power systems is crucial for the optimized operation and precise management of power grids. Busbars are of various types, covering a variety of typical industrial loads as well as common classified loads such as residents and public utilities. Traditional load structure analysis methods mainly rely on manual labeling and expert experience. This method is relatively accurate when dealing with buses with a single industrial type of load, but when faced with buses with multiple types of industrial loads, its accuracy and efficiency are significantly reduced. In addition, traditional methods cannot effectively cope with the dynamic changes in load types caused by changes in enterprise production, and it is difficult to achieve real-time updates and accurate identification of load types.

[0003] Although the prior art has made some progress, such as the bus load prediction method based on user-side classification described in the existing patent CN110689162A, which improves the rationality of load prediction by constructing a feature vector set and using a convolutional neural network combined with a GRU neural network for load prediction; the power load identification method and system proposed in the existing patent CN117458511A, which collects current and voltage information on the power bus, performs load event detection and steady-state current harmonic analysis, and uses a trained model for identification processing, thereby achieving optimal control of electrical appliances on the power bus. However, these methods either focus on prediction rather than real-time identification, or do not fully utilize the advantages of the dynamic time warping (DTW) algorithm in load type identification, especially when processing load data with time series characteristics.

[0004] The prior art lacks a method that can accurately and efficiently identify and update composite load types in real time for dynamic identification of bus load types and is unable to update them in a timely manner. Especially in scenarios where multiple types of industrial loads are combined and load types change dynamically, there is an urgent need for an intelligent identification method, system and equipment for power system load types. Summary of the invention

[0005] The purpose of the present invention is to provide a method, system and device for intelligently identifying load types in a power system, so as to solve the problem of lack of dynamic identification and change of bus load types and failure to update in time.

[0006] To achieve the above purpose, the following technical solution is adopted.

[0007] A method for intelligently identifying load types of a power system comprises the following steps:

[0008] Determine cluster centers of multiple load type subgroups respectively, each cluster center represents a pre-calibrated typical load type;

[0009] The dynamic time warping (DTW) algorithm is used to calculate the DTW distance between the load node to be identified and each cluster center.

[0010] According to the calculated DTW distance, the load node to be identified is assigned to the subgroup of the nearest cluster center to determine the load type of the load node to be identified;

[0011] Regularly scan and identify changes in load types of load units within the domain, and update load type identification results.

[0012] Optionally, in the step of assigning the load node to be identified to a subgroup of the nearest cluster center according to the calculated DTW distance, and determining the load type of the load node to be identified,

[0013] When the DTW distance difference is less than the set value, the variable sliding window scale technology is used to adjust the size and position of the time window, and multi-scale DTW distance calculation is performed on the load node to be identified to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under different scale windows. If so, the load type of the load node to be identified is determined.

[0014] Optionally, when the DTW distance after the variable sliding window scale technology is still unable to uniquely determine the load type of the load node to be identified, a deep classification step is entered, specifically including:

[0015] The power signal decomposition method is used to decompose the power time series data of the load node to be identified to obtain multiple sub-modal signals;

[0016] For each submodal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding submodal signal of each cluster center;

[0017] Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results;

[0018] Based on the similarity evaluation results of each sub-mode, the load type of the load node to be identified is determined using decision logic or integrated learning model.

[0019] Optionally, the specific steps of comprehensively evaluating the similarity of each sub-mode and using decision logic or integrated learning model to determine the load type of the load node to be identified include:

[0020] Assign weights to the similarity evaluation results of each sub-modality to obtain a weighted similarity score, where the weights are determined based on signal stability, historical data performance, or expert input;

[0021] Using weighted similarity scores, the overall similarity of each cluster center is evaluated through decision logic, which includes setting thresholds, comparing scores, and applying logical rules;

[0022] When the weighted similarity score of a single cluster center is significantly higher than that of other cluster centers, the load type of the load node to be identified is determined to be the load type represented by the single cluster center;

[0023] If the weighted similarity score difference between multiple cluster centers is less than the set value and cannot be clearly distinguished, the pre-trained ensemble learning model is used to make the final load type determination based on the sub-modal similarity evaluation results of the load node to be identified.

[0024] Optional, specific calculation method of DTW distance, including

[0025]

[0026] Where: A is the typical load type calibrated in advance, B is the load node to be identified, a i is the load power data of A at time i, b j is the load power data of B at time j;

[0027]

[0028] Where: d(a i , b j ) represents the time series a i and b j The Euclidean distance, DTW distance D(a i , b j ) is the sum of the minimum dynamic distances between A and B at the current moment and adjacent moments, including the cumulative distance of data at past time points.

[0029] Optionally, a dynamic distance matrix H = (A, B) of the clustering index of the two load units is constructed based on the DTW distance calculated in (2) as an element;

[0030]

[0031] According to H = (A, B), an optimal path P is solved to minimize the cumulative loss of the nonlinear mapping of the two-unit power curve clustering. The optimal path P is p 1 = (1, 1) as the starting point, with p k =(i, j) is the end point;

[0032] The DTW distance is

[0033] An intelligent identification system for load types of power systems, comprising:

[0034] A cluster center determination module, used to respectively determine the cluster centers of multiple load type subgroups, each cluster center represents a pre-calibrated typical load type;

[0035] The DTW distance calculation module is used to calculate the DTW distance between the load node to be identified and each cluster center using the dynamic time warping DTW algorithm;

[0036] A load type allocation module is used to allocate the load node to be identified to the subgroup of the nearest cluster center according to the calculated DTW distance, and determine the load type of the load node to be identified;

[0037] The periodic update module is used to periodically scan and identify changes in load types of load units within the domain and update load type identification results.

[0038] Optionally, it also includes a variable sliding window scale processing module, which is used to use a variable sliding window scale technology to adjust the size and position of the time window when the DTW distance difference is less than a set value, and perform multi-scale DTW distance calculation on the load node to be identified, to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under different scale windows, and if so, determine the load type of the load node to be identified.

[0039] Optionally, it also includes a deep classification processing module, which is used to decompose the power time series data of the load node to be identified by using power signal decomposition and decomposition means to obtain multiple sub-modal signals;

[0040] For each submodal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding submodal signal of each cluster center;

[0041] Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results;

[0042] Based on the similarity evaluation results of each sub-mode, the load type of the load node to be identified is determined using decision logic or integrated learning model.

[0043] An electronic device comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement a method for intelligently identifying load types of a power system.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention adopts the dynamic time warping DTW algorithm, which can effectively deal with the deformation problem of time series data and improve the identification accuracy of multi-type industrial load composite busbars. The DTW algorithm takes into account the global characteristics of time series data and can capture the deformation of load power data such as translation and expansion on the time axis, thereby reducing misjudgment and deviation. It can realize dynamic identification and timely update of busbar load types.

[0046] The present invention realizes intelligent adaptive identification of load types, can adapt to changes in load characteristics caused by changes in enterprise production, timely update the identification of load types, and enhance the adaptability and response speed of the power system to changes.

[0047] The present invention avoids the time-consuming manual labeling and expert experience judgment in the traditional method, and significantly improves the efficiency of load type identification through an automated calculation process.

[0048] The present invention regularly scans and identifies changes in load types of load units within a domain, thereby ensuring the timeliness of load type identification results and making power system management more accurate and efficient.

[0049] The present invention further enhances the method's ability to identify load changes on different time scales through the variable sliding window scale technology, ensuring that changes in load characteristics in different time periods can be accurately captured.

[0050] The deep classification step of the present invention provides a more detailed and in-depth load type identification through the application of submodal signal decomposition and integrated learning model, and solves the classification ambiguity problem of similarity.

[0051] The present invention provides a complete systematic solution, including modules such as cluster center determination, DTW distance calculation, load type allocation, variable sliding window scale processing and deep classification processing, forming an efficient collaborative system and further improving the overall performance of load type identification.

[0052] The electronic device of the present invention executes the computer program in the memory, so that the method can be widely deployed in different power systems, and has good universality and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of an embodiment of a method for intelligently identifying load types of a power system according to the present invention.

[0054] Figure 2 It is a flow chart of Example 8 of an intelligent identification method for load types in a power system according to the present invention.

[0055] Figure 3It is a module schematic diagram of an electronic device embodiment according to the present invention.

[0056] Figure 4 It is a schematic diagram of the principle of the variable sliding window scale technology according to an embodiment of the method for intelligent identification of load types in a power system of the present invention.

[0057] Among them, 100, electronic device; 101, memory; 102, processor; 103, computer program; 104, communication bus. DETAILED DESCRIPTION

[0058] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0059] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0060] Example 1

[0061] like Figure 1 As shown, a method for intelligently identifying load types of a power system comprises the following steps:

[0062] Select representative standard load data that reflects the characteristic patterns of different load types. Use unsupervised learning algorithms, such as K-means clustering algorithm, to analyze the standard load data and determine K cluster centers, each of which represents a specific load type.

[0063] For each load node to be identified, its power time series data is collected. The dynamic time warping (DTW) algorithm is used to calculate the DTW distance between the power time series data of the load node to be identified and the representative time series data of each cluster center.

[0064] The power time series data of each load node to be identified is compared with the DTW distance of each cluster center. According to the comparison result, the load node to be identified is assigned to the subgroup represented by the cluster center with the smallest DTW distance, so as to determine the load type of the load node.

[0065] Set a periodic time interval to rescan and identify the load types of all load units in the domain. Update the load type identification results of each load node based on the latest load data to ensure that the load type identification results reflect the current power system status.

[0066] During the implementation process, the parameters of the K-means algorithm can be adjusted according to the characteristics of the power systems in different regions to adapt to power systems of different sizes and types. The implementation of the DTW algorithm can adopt a variety of optimization techniques, such as reducing the search space and adopting fast approximation algorithms to improve computational efficiency. In the load type allocation step, a confidence assessment mechanism can be introduced. When the DTW distance difference is small, the load type is further determined through confidence assessment. The regular update mechanism can be dynamically adjusted according to the operating status of the power system and the frequency of load changes to achieve more accurate load type management.

[0067] Through the above implementation, the method can effectively and intelligently identify the load type in the power system, improve the automation level and accuracy of load management, and at the same time have good adaptability and scalability.

[0068] Example 2

[0069] like Figure 2 and Figure 4 As shown, as a preferred example, a preliminary calculation of the DTW distance between each load node to be identified and each cluster center is completed to determine the preliminary load type of each node. When the DTW distance between the load node to be identified and multiple cluster centers is close, and the gap is less than the set value, and its load type cannot be clearly determined, the variable sliding window scale technology is introduced for further analysis. Adjust the size and position of the time window to adapt to the changes in load characteristics in different time periods. The adjustment of the time window is based on the power time series data of the load node to be identified.

[0070] Under the adjusted time window, the DTW distance between the load node to be identified and each cluster center is recalculated to achieve multi-scale similarity evaluation. The DTW distance calculated under different scale windows is analyzed to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under all scales. If the load nodes to be identified have the minimum DTW distance with the same cluster center under multi-scale windows, the load type of the node is determined to be the load type represented by the cluster center.

[0071] The variable sliding window scaling technology can be customized according to the characteristics of the actual power system load data to adapt to different load change patterns and time series characteristics. The adjustment range and step value of a scale window can be set to systematically explore the optimal time window configuration. During the implementation process, the adjustment strategy of the scale window can be dynamically optimized based on historical data and real-time data in combination with expert systems or artificial intelligence algorithms. Automated tools can be further developed to monitor the calculation results of the DTW distance in real time and automatically trigger the application of the variable sliding window scaling technology.

[0072] Through the above implementation, the method can provide more accurate load type identification through the variable sliding window scale technology when the DTW distance gap is less than the set value, thereby enhancing the robustness and adaptability of the method.

[0073] Example 3

[0074] As a preferred example, after the variable sliding window scale technology is processed, if the DTW distance of the load node to be identified still cannot uniquely determine its load type, the deep classification step is triggered. The power time series data of the load node to be identified is decomposed and decomposed, and the original signal is decomposed into multiple submodal signals by Fourier transform, wavelet transform or other suitable signal processing technology. For each submodal signal, the DTW algorithm is applied independently, and the DTW distance is calculated with the corresponding submodal signal of each cluster center to evaluate the similarity between each submodal signal and each load type. Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results. The DTW distance and similarity evaluation results of each submodal signal are combined, and the overall similarity of each cluster center is evaluated by weighted average, voting mechanism or other decision logic methods. Decision logic is formulated, including setting thresholds for similarity scores, comparing scores, applying logical rules, etc., to determine the load type of the load node to be identified.

[0075] If the similarity score of a single cluster center does not have a significant advantage, a pre-trained ensemble learning model, such as a random forest, gradient boosting tree, or neural network, is used to conduct a comprehensive analysis and final judgment on the submodal similarity evaluation results. The ensemble learning model is trained and validated, using historical load data as a training set, and the generalization ability of the model is evaluated through cross-validation and other techniques.

[0076] The deep classification steps can be adjusted according to the specific power system characteristics and load characteristics to adapt to different application scenarios and needs. The signal decomposition and decomposition technology can be further optimized to select a decomposition method that is more suitable for specific types of load data to improve the recognition accuracy of submodal signals. In the implementation process, expert knowledge and artificial intelligence algorithms can be combined to dynamically adjust the weight distribution of submodal signals and optimize decision logic. The integrated learning model can be designed as an adaptive learning system to continuously optimize model parameters based on new data and feedback to improve recognition accuracy.

[0077] Through the above implementation, the method can provide a more accurate and reliable load type identification result through the deep classification step when the DTW distance cannot uniquely determine the load type, thereby enhancing the accuracy and robustness of the method.

[0078] Example 4

[0079] As a preferred example, the specific steps of comprehensively evaluating the similarity of each sub-mode and using decision logic or integrated learning model to determine the load type of the load node to be identified include:

[0080] Assign weights to the similarity evaluation results of each sub-modality to obtain a weighted similarity score, where the weights are determined based on signal stability, historical data performance, or expert input;

[0081] Using weighted similarity scores, the overall similarity of each cluster center is evaluated through decision logic, which includes setting thresholds, comparing scores, and applying logical rules;

[0082] When the weighted similarity score of a single cluster center is significantly higher than that of other cluster centers, the load type of the load node to be identified is determined as the load type represented by the single cluster center;

[0083] If the weighted similarity scores of multiple cluster centers are close and the difference is less than the set value, they cannot be clearly distinguished. Then, the pre-trained ensemble learning model is used to make the final load type determination based on the sub-modal similarity evaluation results of the load node to be identified.

[0084] Specifically, for each submodal signal, a corresponding weight is assigned to the submodal similarity evaluation result according to its strength of association with the load type, stability and historical data performance. These weights can be input through the expert system or determined based on statistical analysis. Calculate the weighted similarity score for each submodal signal. This score reflects the similarity between the submodal signal and each cluster center and takes into account the influence of the weight. Develop a decision logic, including setting a threshold for the weighted similarity score and comparing the scores of different cluster centers. The decision logic also includes applying logical rules, such as giving priority to the cluster center with the highest weighted score. Using the weighted similarity score, the overall similarity of each cluster center is evaluated. This evaluation process may include comparing, sorting and filtering the scores.

[0085] If the weighted similarity score of a cluster center is significantly higher than that of other cluster centers, the load type of the load node to be identified is determined to be the load type represented by the cluster center. When the weighted similarity scores of multiple cluster centers are similar and cannot be clearly distinguished by decision logic, the pre-trained ensemble learning model is used for final judgment. The ensemble learning model is trained using historical load data and verified by cross-validation and other techniques to ensure the accuracy and generalization ability of the model.

[0086] The weight distribution of weighted similarity scores can be adjusted according to the actual operating characteristics and load characteristics of the power system to improve the accuracy of load type identification. The decision logic can be customized according to the needs of different application scenarios to adapt to different operating environments and decision requirements. The ensemble learning model can use a variety of algorithms, such as random forests, gradient boosting machines, or neural networks, to achieve the best identification effect. Automated tools can be further developed to monitor the weighted similarity scores in real time and automatically trigger the application of ensemble learning models to improve the efficiency of the identification process.

[0087] Through the above implementation, when the difference between the sub-modal similarity evaluation results is less than the set value, the method can provide accurate load type identification through the application of comprehensive evaluation and integrated learning model, thereby enhancing the reliability and practicality of the method.

[0088] Example 5

[0089] As a preferred example, the specific calculation method of DTW distance includes:

[0090]

[0091] Where: A is the typical load type calibrated in advance, B is the load node to be identified, a i is the load power data of A at time i, b j is the load power data of B at time j;

[0092]

[0093] Where: d(a i , b j ) represents the time series a i and b j The Euclidean distance, DTW distance D(a i , b j ) is the sum of the minimum dynamic distances between A and B at the current moment and adjacent moments, including the cumulative distance of data at past time points.

[0094] Specifically, define two time series A = a 1 ,a 2 ,...,a i and B = b 1 ,b 2 ,...,b j , where A represents the load power data of the pre-calibrated typical load type at different time points, and B represents the load power data of the load node to be identified at the same time point.

[0095] Initialize a two-dimensional matrix D to store all possible DTW distances between sequences A and B.

[0096] According to formula (2), when i = 1 and j = 1, calculate d(a 1 ,b 1 ) as the DTW distance in the initial state, where d(a i ,b j ) means a i and b j The Euclidean distance between two points.

[0097] For all other values ​​of i and j, D(a) is calculated recursively according to formula (2) i ,b j ). This involves considering the minimum dynamic distance between the current moment and the adjacent moments and adding it to the accumulated distance in the past.

[0098] Through recursive calculation, determine a path from (a 1 ,b 1 ) to (a i ,b j ), the cumulative distance of this path is the DTW distance between the two sequences.

[0099] In a specific implementation, the calculation method of the DTW distance may be adjusted according to the actual characteristics of the power system load data, for example, by limiting the search range or using a heuristic algorithm to improve the calculation efficiency.

[0100] Can be d(a i ,b j ) is optimized to adapt to different types of load power data, for example by introducing weights or considering the local characteristics of the data.

[0101] Automated tools can be developed to automate and parallelize the DTW distance calculation process to handle large-scale load data.

[0102] Through the above implementation, the method can accurately calculate the DTW distance between the typical load type and the load node to be identified in the power system, and provides an effective quantitative means for intelligent identification of load types.

[0103] Example 6

[0104] As a preferred example, the dynamic distance matrix H(A, B) of the clustering index of the two load units is constructed based on the DTW distance calculated in (2) as an element;

[0105]

[0106] According to H(A, B), an optimal path P is solved to minimize the cumulative loss of the nonlinear mapping of the two-unit power curve clustering. The optimal path P is p1 =(1,1) as the starting point, with p k =(i, j) is the end point;

[0107] The DTW distance is

[0108] Specifically, the DTW distance calculated by formula (2) is used as an element to construct a dynamic distance matrix H(A, B), where A and B represent the time series data of two load units respectively.

[0109] In the dynamic distance matrix H(A,B), find an optimal path P that starts from the starting point (1,1) and ends at the end point (i,j), where i and j are the lengths of the two time series respectively.

[0110] The optimal path P minimizes the cumulative loss of two-unit power curve clustering through nonlinear mapping, that is, minimizes the sum of all DTW distances on the path.

[0111] DTW distance is defined as the sum of all elements on the optimal path P, that is, DTW(A,B)=∑ pi∈P D(p i ), where D(p i ) is the DTW distance of the i-th element on path P.

[0112] The construction of the dynamic distance matrix H(A,B) can adopt known algorithms, such as dynamic programming, to ensure computational efficiency and accuracy. The solution of the optimal path P can be achieved through graph search algorithms, such as the Dijkstra algorithm or the Floyd-Warshall algorithm, which can find the shortest path from the starting point to the end point in the matrix. The process of minimizing the cumulative loss can be adjusted in combination with the actual load characteristics of the power system to adapt to different load patterns and changing trends. The calculation results of the DTW distance can be used for cluster analysis to help identify load units with similar load patterns and provide a basis for the optimal management of the power system. The DTW distance DTW(A,B) can represent the dynamic similarity between two load units. The smaller the DTW value, the higher the dynamic similarity between the two load units. The more it can be regarded as the same type of load.

[0113] Through the above implementation, the method can effectively use the DTW distance to intelligently identify the load type of the power system, thereby improving the accuracy and efficiency of load management.

[0114] Example 7

[0115] An intelligent identification system for load types of power systems, comprising:

[0116] A cluster center determination module, used to respectively determine the cluster centers of multiple load type subgroups, each cluster center represents a pre-calibrated typical load type;

[0117] The DTW distance calculation module is used to calculate the DTW distance between the load node to be identified and each cluster center using the dynamic time warping DTW algorithm;

[0118] A load type allocation module is used to allocate the load node to be identified to the subgroup of the nearest cluster center according to the calculated DTW distance, and determine the load type of the load node to be identified;

[0119] The periodic update module is used to periodically scan and identify changes in load types of load units within the domain and update load type identification results.

[0120] Specifically, this module is responsible for determining the representative of each typical load type, i.e., the cluster center, from the historical load data through cluster analysis. These cluster centers are automatically selected based on specific load characteristics, such as power consumption mode, time distribution, etc., through algorithms such as the K-means algorithm. This module uses the dynamic time warping (DTW) algorithm to calculate the DTW distance between the time series data of the load node to be identified and the time series data of each cluster center. This module includes steps such as data preprocessing, distance matrix initialization, recursive calculation of the optimal path and distance accumulation. Based on the results of the DTW distance calculation module, this module assigns each load node to be identified to the subgroup represented by the cluster center with the closest DTW distance to it. This process involves operations such as distance comparison, subgroup selection and load type marking. This module sets a periodic time interval to automatically trigger the rescanning and identification of the load types of all load units in the domain. This module includes steps such as data collection, real-time monitoring, load type re-identification and database update.

[0121] The cluster center determination module can further include preprocessing steps such as data cleaning, feature selection and multidimensional scaling to improve the accuracy and efficiency of cluster analysis. The DTW distance calculation module can use optimization algorithms, such as fast DTW algorithm, to reduce calculation time and improve system response speed. The load type allocation module can introduce a confidence assessment mechanism to conduct a more detailed analysis of the situation where the DTW distance gap is less than the set value to improve the accuracy of classification. The regular update module can dynamically adjust the update frequency and range based on the actual operation of the power system, such as peak load period.

[0122] This system can realize efficient and intelligent identification of power system load types, and can adapt to dynamic changes in load characteristics, providing strong technical support for the optimal dispatch and management of the power system.

[0123] Example 8

[0124] As a preferred example, it also includes a variable sliding window scale processing module, which is used to use a variable sliding window scale technology to adjust the size and position of the time window when the DTW distance difference is less than a set value, and perform multi-scale DTW distance calculation on the load node to be identified, to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under different scale windows, and if so, determine the load type of the load node to be identified.

[0125] Specifically, in the load type intelligent identification system, a variable sliding window scale processing module is integrated to handle the situation where the DTW distance difference is less than the set value, thereby improving the accuracy of load type identification. When the DTW distance calculation module in the system obtains a similar DTW distance result, the variable sliding window scale processing module is automatically triggered to adjust the size and position of the time window to adapt to the load characteristics of different time periods. The variable sliding window scale processing module performs multi-scale analysis on the time series data of the load node to be identified, and calculates the DTW distance at different time scales to obtain more comprehensive load characteristic information. Under each adjusted time window, the module recalculates the DTW distance between the load node to be identified and each cluster center, and compares these distances to determine whether there is a DTW distance with the minimum consistency. If the DTW distance between the load node to be identified and a cluster center is always the smallest under multiple scale windows, the variable sliding window scale processing module determines that the load type of the node is the type represented by the cluster center.

[0126] The variable sliding window scale processing module can set different time window adjustment strategies according to the actual characteristics of the power load data, such as the increase or decrease ratio of the window size, the sliding step size, etc. The module can combine historical data and real-time data, and automatically optimize the adjustment parameters of the time window through machine learning methods to improve the intelligence level of the identification system. In order to enhance the adaptability of the system, the variable sliding window scale processing module can be designed to be configurable, allowing users to customize it according to different power system environments and load change trends. The implementation of the module can adopt an efficient algorithm to process large-scale data and ensure the real-time response capability of the system. Through the above implementation method, the system can effectively solve the problem of load type identification when the DTW distance is close, and improve the accuracy and reliability of intelligent identification of power system load types.

[0127] Example 9

[0128] As a preferred example, it also includes a deep classification processing module, which is used to decompose the power time series data of the load node to be identified by using a power signal decomposition method to obtain multiple sub-modal signals;

[0129] For each submodal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding submodal signal of each cluster center;

[0130] Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results;

[0131] Based on the similarity evaluation results of each sub-mode, the load type of the load node to be identified is determined using decision logic or integrated learning model.

[0132] Specifically, in the intelligent identification system, a deep classification processing module is integrated to further refine the identification of the load type when the variable sliding window scale processing module cannot uniquely determine the load type. The deep classification processing module uses power signal decomposition and decomposition methods, such as wavelet transform or Fourier transform, to decompose the power time series data of the load node to be identified into multiple sub-modal signals. For each sub-modal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding sub-modal signal of each pre-calibrated cluster center to evaluate the similarity of each sub-modality. Based on the DTW distance of each sub-modal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of sub-modal similarity evaluation results. The deep classification processing module combines the similarity evaluation results of each sub-modal signal and uses decision logic or integrated learning model to make the final load type determination. Formulate a clear decision logic, including determining weight allocation, similarity threshold setting, score comparison, etc., to select the cluster center with the highest weighted similarity score as the load type.

[0133] If the submodal similarity evaluation results are similar, that is, the difference is less than the set value and cannot be clearly distinguished, an integrated learning model, such as a random forest or a gradient boosting machine, is used to comprehensively analyze the evaluation results and make a final judgment. If the submodal similarity evaluation results are similar, that is, the difference is less than the set value and cannot be clearly distinguished, an integrated learning model, such as a random forest or a gradient boosting machine, is used to comprehensively analyze the evaluation results and make a final judgment. The integrated learning model is trained and validated, using labeled load data as the training set, and the generalization ability of the model is evaluated through techniques such as cross-validation.

[0134] The signal decomposition means of the deep classification processing module can be selected and optimized according to different types of load characteristics to improve the decomposition effect and identification accuracy. The calculation of the DTW algorithm can adopt a variety of optimization techniques, such as reducing the search space and using fast approximation algorithms to improve computational efficiency. The decision logic and integrated learning model can be designed as an adaptive learning system to continuously optimize parameters based on new data and feedback to improve identification accuracy. The system can further develop automated tools to monitor submodal similarity scores in real time and automatically trigger the application of integrated learning models to improve the efficiency of the identification process.

[0135] Through the above implementation, when the load type still cannot be determined after the variable sliding window scale processing, the system can provide a more accurate and reliable load type identification result through the deep classification processing module.

[0136] Example 10

[0137] like Figure 3 As shown, an electronic device includes a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement an intelligent identification method for load types of a power system.

[0138] The present invention also provides an electronic device 100 for implementing a method for intelligently identifying load types in a power system according to the above embodiment; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104. The memory 101 can be used to store the computer program 103, and the processor 102 implements the steps of a method for intelligently identifying load types in a power system by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or the processor 102 may also be any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100. The memory 101 in the electronic device 100 stores multiple instructions to implement a method for intelligent identification of load types in a power system.

[0139] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

Claims

1. A method for intelligent identification of load types in a power system, characterized in that: The following steps are included: Determine cluster centers of multiple load type subgroups respectively, each cluster center represents a pre-calibrated typical load type; The dynamic time warping (DTW) algorithm is used to calculate the DTW distance between the load node to be identified and each cluster center. According to the calculated DTW distance, the load node to be identified is assigned to the subgroup of the nearest cluster center to determine the load type of the load node to be identified; Regularly scan and identify changes in load types of load units within the domain and update load type identification results.

2. The method for intelligent identification of load types in a power system according to claim 1, characterized in that: According to the calculated DTW distance, the load node to be identified is assigned to the subgroup of the nearest cluster center, and in the step of determining the load type of the load node to be identified, When the DTW distance difference is less than the set value, the variable sliding window scale technology is used to adjust the size and position of the time window, and multi-scale DTW distance calculation is performed on the load node to be identified to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under different scale windows. If so, the load type of the load node to be identified is determined.

3. The method for intelligent identification of load types in a power system according to claim 2, characterized in that: When the DTW distance after the variable sliding window scale technology is still unable to uniquely determine the load type of the load node to be identified, the deep classification step is entered, which includes: The power signal decomposition method is used to decompose the power time series data of the load node to be identified to obtain multiple sub-modal signals; For each submodal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding submodal signal of each cluster center; Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results; Based on the similarity evaluation results of each sub-mode, the load type of the load node to be identified is determined using decision logic or integrated learning model.

4. The method for intelligent identification of load types in a power system according to claim 3, characterized in that: Based on the similarity evaluation results of each sub-mode, the specific steps of determining the load type of the load node to be identified by using decision logic or integrated learning model include: Assign weights to the similarity evaluation results of each sub-modality to obtain a weighted similarity score, where the weights are determined based on signal stability, historical data performance, or expert input; Using weighted similarity scores, the overall similarity of each cluster center is evaluated through decision logic, which includes setting thresholds, comparing scores, and applying logical rules; When the weighted similarity score of a single cluster center is higher than that of other cluster centers, the load type of the load node to be identified is determined to be the load type represented by the single cluster center; If the weighted similarity score difference between multiple cluster centers is less than the set value and cannot be distinguished, the pre-trained ensemble learning model is used to make the final load type determination based on the sub-modal similarity evaluation results of the load node to be identified.

5. The method for intelligent identification of load types in a power system according to claim 1, characterized in that: The specific calculation method of DTW distance includes Where: A is the typical load type calibrated in advance, B is the load node to be identified, a i is the load power data of A at time i, b j is the load power data of B at time j; Where: d(a i , b j ) represents the time series a i and b j The Euclidean distance, DTW distance D(a i , b j ) is the sum of the minimum dynamic distances between A and B at the current moment and adjacent moments, including the cumulative distance of data at past time points.

6. A method for intelligent identification of load types in a power system according to claim 5, characterized in that: Based on the DTW distance calculated by (2) as the element, the dynamic distance matrix H(A, B) of the clustering index of the two load units is constructed; According to H(A, B), an optimal path P is solved to minimize the cumulative loss of the nonlinear mapping of the two-unit power curve clustering. The optimal path P starts with p1=(1, 1) and ends with p k =(i, j) is the end point; The DTW distance is 7. An intelligent identification system for load types of a power system, based on an intelligent identification method for load types of a power system according to any one of claims 1 to 6, characterized in that: include, A cluster center determination module, used to respectively determine the cluster centers of multiple load type subgroups, each cluster center represents a pre-calibrated typical load type; The DTW distance calculation module is used to calculate the DTW distance between the load node to be identified and each cluster center using the dynamic time warping DTW algorithm; A load type allocation module is used to allocate the load node to be identified to the subgroup of the nearest cluster center according to the calculated DTW distance, and determine the load type of the load node to be identified; The periodic update module is used to periodically scan and identify changes in load types of load units within the domain and update load type identification results.

8. The intelligent identification system for load types of a power system according to claim 7, characterized in that: Also includes, The variable sliding window scale processing module is used to use the variable sliding window scale technology to adjust the size and position of the time window when the DTW distance difference is less than the set value, and perform multi-scale DTW distance calculation on the load node to be identified to determine whether the load node to be identified shows the minimum DTW distance with a certain cluster center under different scale windows. If so, the load type of the load node to be identified is determined.

9. The intelligent identification system for load types of a power system according to claim 8, characterized in that: It also includes a deep classification processing module, which is used to decompose the power time series data of the load node to be identified by using power signal decomposition and decomposition means to obtain multiple sub-modal signals; For each submodal signal, the DTW algorithm is applied independently to calculate the DTW distance with the corresponding submodal signal of each cluster center; Based on the DTW distance of each submodal signal, the similarity between the load node to be identified and each cluster center is evaluated to obtain a set of submodal similarity evaluation results; Based on the similarity evaluation results of each sub-mode, the load type of the load node to be identified is determined using decision logic or integrated learning model.

10. An electronic device, characterized in that: It comprises a processor (102) and a memory (101), wherein the processor (102) is used to execute a computer program (103) stored in the memory (101) to implement a method for intelligently identifying load types in a power system as claimed in any one of claims 1 to 6.

Images