A method for remotely controlling power data

Abstract

The present invention relates to the field of power system monitoring and control technology, and provides a method for remotely controlling power data, including a data acquisition module and a processor: the data acquisition module acquires power operation data; the processor inputs the data into a power load analysis network to obtain a load classification result, and accordingly acquires the equipment current waveform data and operation status data, and generates a power state vector representing the influence of parameter changes on the waveform amplitude. The influence weights of the state parameters on each section of the current waveform are determined by analyzing the network, and multiple analysis dimensions are obtained, and the equipment operation status data is divided into multiple subsets containing waveform data accordingly. Finally, the state evaluation network is used to determine whether the equipment state is abnormal based on the divided data group and generate a control instruction. The present invention can realize the accurate identification of load types and abnormal states, and improve the automation level and reliability of remote control of power systems.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring and control, and more specifically, to a method for remotely managing and controlling power data. Background Art

[0002] In today's society, the stable operation of power systems plays a vital role in all areas of society. With the continuous development and complexity of power systems, the demand for power data management and control is also increasing. Traditional power data management and control methods mainly rely on on-site manual monitoring and simple automated equipment, which has many limitations. First, manual monitoring requires a lot of manpower and time and is easily affected by human factors, making it difficult to ensure the accuracy and timeliness of data. Second, traditional automated equipment has relatively limited functions and is unable to conduct comprehensive and in-depth analysis and processing of power data, making it difficult to meet the high-precision and high-efficiency data management requirements of modern power systems.

[0003] In recent years, with the rapid development of information technology, a number of emerging technologies have been introduced into the field of power data management and control. For example, power operation data can be collected through sensor networks and preliminary data processing can be performed using computer algorithms. These technologies have improved the efficiency and accuracy of power data management and control to a certain extent, but some problems still exist. On the one hand, existing data collection and processing methods can often only obtain limited power operation data, such as basic parameters such as voltage and current, but lack effective means to obtain and analyze in-depth information on the operating status of equipment, such as the correlation between equipment current waveform data and equipment operating status data. On the other hand, when processing complex power data, existing analysis methods have difficulty accurately identifying abnormal equipment conditions and are unable to generate effective control instructions in a timely manner, thus affecting the stable operation of the power system.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technology: the existing technology is unable to comprehensively and accurately obtain and analyze the deep-level information of the operating status of power equipment, and it is difficult to effectively identify the abnormal status of the equipment and generate accurate control instructions, resulting in the efficiency and accuracy of power data control unable to meet the needs of modern power systems. Summary of the Invention

[0005] The present invention provides a method for remotely controlling power data, which is applied to a remote control system. The remote control system includes a data acquisition module and a processor:

[0006] The data acquisition module is used to obtain power operation data; the power operation data includes voltage value, current value and power factor;

[0007] The processor is configured to perform the following steps:

[0008] Inputting the power operation data into a power load analysis network to obtain a plurality of load classification results; the load classification results indicate the load types corresponding to the power equipment;

[0009] According to the load classification result, device current waveform data and corresponding device operating status data are obtained; the device current waveform data represents a periodic current variation curve generated when the power equipment is running; the device operating status data represents a set of device operating parameters corresponding to when the device current waveform data is recorded;

[0010] generating a power state vector based on the device operating state data and the device current waveform data; wherein the power state vector represents the effect of parameter changes in the device operating state data on the amplitude of the current waveform;

[0011] By analyzing the network, based on the device operating status data, determining the influence weight of the status parameter on each section of the device current waveform data, and obtaining multiple analysis dimensions;

[0012] Based on the analysis dimension and the device operating status data, the device operating status data is divided into different dimensions to obtain a plurality of divided data groups; the divided data groups are sub-data sets including the device current waveform data;

[0013] Through the status assessment network, based on the divided data groups, it is determined whether the equipment operation status is abnormal and a control instruction is generated.

[0014] Furthermore, the network is analyzed to determine the influence weight of the state parameter on each segment of the device current waveform data based on the device operating status data, thereby obtaining multiple analysis dimensions, including:

[0015] Inputting the equipment operation status data into an anomaly detection network to identify equipment anomaly characteristics and obtain a first state feature set; the first state feature set represents the degree of deviation of key parameters of the equipment;

[0016] Inputting the first state feature set into a position calibration network to obtain a device parameter association diagram; the device parameter association diagram is a topological diagram that marks multiple parameter influence areas;

[0017] Input the device parameter association diagram into the analysis network to obtain multiple analysis dimensions.

[0018] Furthermore, the training method of the analysis network includes:

[0019] Acquire multiple training data sets; the training data sets are historical device parameter association diagrams;

[0020] Obtaining a power state vector corresponding to the training data set as labeled data;

[0021] Inputting the training data set into the analysis network to obtain a training analysis dimension;

[0022] Based on the training analysis dimension and the labeled data, the training data set is divided to obtain a plurality of training partitioned data groups and corresponding core nodes; the training partitioned data groups are data subsets divided according to the training analysis dimension; the core nodes represent key positions of the first partitioned data group on the labeled reference curve;

[0023] Based on the training partition data group and the core node, the parameter influence degree is calculated with the main path as the benchmark to obtain multiple training partition load total values ​​and corresponding data intervals; the data interval represents the coverage range of the parameter influence area in the training partition data group; the training partition load total value represents the amplitude change of each parameter in the training partition data group during the data interval;

[0024] Calculating a parameter influence index based on a plurality of training partition load total values ​​and corresponding data intervals; wherein the plurality of training partition load total values ​​correspond to one parameter influence index; and wherein a plurality of training partition data groups correspond to obtaining a plurality of parameter influence indexes; wherein the parameter influence index represents the intensity of influence of the parameters of the entire training partition data group on the current waveform;

[0025] Dividing the multiple parameter influence indicators by the sum of the multiple parameter influence indicators to obtain an influence weight coefficient; the influence weight coefficient represents the proportion of the influence of the multiple training divided data groups on the current waveform;

[0026] A loss function is calculated based on the impact weight coefficient and the power state vector to optimize the analysis network.

[0027] Furthermore, generating a power state vector based on the device operating state data and the device current waveform data includes:

[0028] Using the correlation curve between the device operating status data and the device current waveform data as a reference curve;

[0029] Segmenting the device current waveform data along a direction perpendicular to the reference curve to obtain a plurality of first segmented data sets;

[0030] Accumulating the amplitude parameters in the first segmented data set to obtain a segment load value; obtaining multiple segment load values ​​corresponding to multiple first segmented data sets;

[0031] The plurality of segment load values ​​are arranged in sequence according to the time sequence positions of the plurality of first segment data sets in the device current waveform data to generate a power state vector.

[0032] Furthermore, the training data set is divided based on the training analysis dimension and the labeled data to obtain multiple training divided data groups and corresponding core nodes, including:

[0033] Using the first segmented data set of the device current waveform data corresponding to the labeled data as a labeled first segmented set;

[0034] Using the reference curve corresponding to the first segment set as the labeled reference curve;

[0035] Marking the first segmented set according to the positions of the labeled reference curve in the training data set to obtain a plurality of first labeled positions;

[0036] The midpoint of two adjacent first-marked positions is taken as the core node;

[0037] Determine the angle difference from the annotated reference curve as the training analysis dimension and the dividing boundary passing through the core node;

[0038] The training data set is divided using the division boundary to obtain a training division data set.

[0039] Furthermore, the parameter influence degree is calculated based on the training partition data group and the core node with the main path as the benchmark to obtain multiple training partition load total values ​​and corresponding data intervals, including:

[0040] Determining a main path of the training partition data group according to the training partition data group and the core node;

[0041] Based on the training partition data set and the main path, the parameter influence is mapped to the main path to obtain a plurality of training partition load total values; each position on the main path corresponds to a training partition load total value;

[0042] Get the coordinates of the positions on the main path in the training dataset as the division positions;

[0043] According to the coordinates of the partitioned position and the core node, the distance value is calculated using the Euclidean distance formula to obtain the data interval;

[0044] The training partition load total value is associated with the corresponding data interval and stored.

[0045] Furthermore, based on the training partition data group and the main path, the parameter influence is mapped to the main path to obtain a plurality of training partition load total values, including:

[0046] Selecting discrete positions on the main path as partition reference positions; multiple partition reference positions are evenly distributed along the main path;

[0047] Determine a plane passing through the division reference position and perpendicular to the main path as the division reference plane;

[0048] In the training partition data group, the amplitude parameters on the partition reference plane are accumulated to obtain the total training partition load value; multiple partition reference planes correspond to multiple training partition load total values.

[0049] Furthermore, the parameter impact index is calculated based on the total values ​​of the multiple training division loads and the corresponding data intervals, including:

[0050] Sum multiple data intervals to get the interval sum;

[0051] Divide the individual data interval by the sum of the intervals to obtain the interval influence coefficient;

[0052] Multiplying the total value of the training division load by the corresponding interval influence coefficient to obtain a division influence value;

[0053] All the partition influence values ​​are summed up to obtain the parameter influence index; multiple training partition data groups correspond to multiple parameter influence indexes.

[0054] Furthermore, determining the main path of the training divided data group according to the training divided data group and the core node includes:

[0055] If the training divided data group is in a regular form, obtaining key feature points;

[0056] Connect two specific key feature points to generate a reference line;

[0057] Determine key locations on the reference line;

[0058] Associate core nodes with key locations to determine the main path.

[0059] Furthermore, determining the main path of the training divided data group according to the training divided data group and the core node includes:

[0060] If the training data group is irregular, obtain the feature identification points;

[0061] The feature identification points containing the core nodes are used as marking feature points;

[0062] The associated area of ​​the marked feature points is used as the reference area;

[0063] Identify key locations in the reference area;

[0064] Connect key locations and marked feature points to determine the main path.

[0065] According to the above-mentioned embodiment of the present invention, there are at least the following beneficial effects: the method of remotely controlling power data of the present invention can realize comprehensive collection and in-depth analysis of power operation data. A variety of power operation data including voltage value, current value and power factor are obtained through the data acquisition module, and the data is classified and processed with the help of the power load analysis network, which can accurately identify the load type of the power equipment. Furthermore, a power state vector is generated based on the equipment current waveform data and the equipment operation status data, and the influence weight of the state parameter on the current waveform is determined through the analysis network, thereby obtaining multiple analysis dimensions and realizing a refined division of the equipment operation status. This method can effectively improve the accuracy of power data control and provide strong support for the stable operation of the power system.

[0066] Furthermore, the present invention can also use the state assessment network to determine whether the equipment's operating status is abnormal based on the divided data groups and generate corresponding control instructions. This method not only enables timely detection of potential equipment failures but also generates precise control instructions based on the equipment's actual operating status, thereby enabling real-time, remote control of power equipment. This invention can improve the operating efficiency and reliability of power systems, reduce the risks and losses caused by equipment failures, and provide a new technical means for intelligent management of power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0068] Figure 1 A flowchart of a method for remotely managing power data provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0069] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0070] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0071] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0072] Reference below Figure 1 , Figure 1 This is a flow chart of a method for remotely controlling power data provided by an embodiment of the present invention. Figure 1 As shown, a method for remotely controlling power data is applied to a remote control system, which includes a data acquisition module and a processor:

[0073] The data acquisition module is used to obtain power operation data; the power operation data includes voltage value, current value and power factor;

[0074] The processor is configured to perform the following steps:

[0075] S1 inputs the power operation data into the power load analysis network to obtain multiple load classification results; the load classification results represent the load types corresponding to the power equipment;

[0076] S2, based on the load classification result, obtains device current waveform data and corresponding device operating status data; the device current waveform data represents a periodic current variation curve generated when the power device is operating; the device operating status data represents a set of device operating parameters corresponding to when recording the device current waveform data;

[0077] S3 generates a power state vector based on the device operating state data and the device current waveform data; the power state vector represents the effect of parameter changes in the device operating state data on the amplitude of the current waveform;

[0078] S4 determines the influence weight of the state parameter on each segment of the device current waveform data by analyzing the network based on the device operating state data, thereby obtaining multiple analysis dimensions;

[0079] S5: dividing the device operating status data into different dimensions based on the analysis dimension and the device operating status data to obtain a plurality of divided data groups; the divided data groups are sub-data sets including the device current waveform data;

[0080] S6 determines whether the equipment operation status is abnormal based on the divided data group through the status assessment network and generates a control instruction.

[0081] It should be noted that the present invention relates to a method for remotely controlling power data, which uses a data acquisition module and a processor to collect, analyze, and control power operation data. The data acquisition module is used to obtain power operation data, which includes key parameters such as voltage, current, and power factor. These parameters are the basic reflection of the operating status of the power system. The processor is responsible for executing a series of analysis and processing steps to achieve accurate assessment and control of the operating status of power equipment. The power load analysis network is a model used to classify and process power operation data. It can output multiple load classification results based on the input power operation data. These results reflect the load type corresponding to the power equipment. Equipment current waveform data refers to the periodic current change curve generated by the power equipment during operation. It can reflect the current characteristics of the equipment under different operating conditions. Equipment operation status data is a set of equipment operating parameters corresponding to the recording of equipment current waveform data. These parameter sets provide basic data support for subsequent analysis. Through these steps, the present invention can achieve comprehensive monitoring and precise control of the operating status of power equipment.

[0082] Specifically, the power operation data collected by the data acquisition module are key indicators of power system operation. Voltage refers to the voltage level in power lines, current refers to the current flowing through power equipment, and power factor is a key parameter for measuring power transmission efficiency in power systems. The power load analysis network analyzes this data to classify power equipment load types into different categories, such as industrial, commercial, or residential loads, enabling targeted analysis and control. Equipment current waveform data reflects the dynamic changes in current during operation. It is a periodic curve that contains rich information about the equipment's operating status. Equipment operating status data includes equipment operating parameters, such as start-up time, operating duration, and load factor. These parameters are correlated with the equipment current waveform data, providing comprehensive data support for subsequent analysis. The analysis network is a model that determines the influence weight of each state parameter on each segment of the equipment current waveform data. It inputs equipment operating status data and outputs multiple analysis dimensions that help further refine the assessment of equipment operating status. The status assessment network is a model that determines whether the equipment operating status is abnormal based on the divided data groups and generates control instructions. It can determine whether the equipment has faults or other abnormal conditions based on the input divided data groups, and generate corresponding control instructions accordingly to achieve remote control of power equipment.

[0083] Preferably, the construction of the power load analysis network can be trained by collecting a large amount of historical power operation data, which includes parameters such as voltage, current, and power factor under different load types. Through machine learning algorithms, such as neural networks or decision tree algorithms, these data are classified and trained to construct a power load analysis network model that can accurately identify different load types. During the data processing process, the device current waveform data can be segmented according to its periodic characteristics. For example, a complete current waveform cycle can be divided into several equal-length time periods, and the current amplitude in each time period is used as the input parameter for analysis. The parameters in the device operation status data can be normalized according to their physical meaning to better correlate and analyze with the current waveform data. For example, the load rate of the device can be normalized to a range of 0 to 1, where 0 indicates that the device is not running and 1 indicates that the device is running at full load. Through these detailed processing steps, the accuracy and reliability of the analysis can be further improved, providing more accurate data support for the remote control of power equipment.

[0084] In some embodiments, the network is analyzed to determine the influence weight of the state parameter on each segment of the device current waveform data based on the device operating state data, thereby obtaining multiple analysis dimensions, including:

[0085] Inputting the equipment operation status data into an anomaly detection network to identify equipment anomaly characteristics and obtain a first state feature set; the first state feature set represents the degree of deviation of key parameters of the equipment;

[0086] Inputting the first state feature set into a position calibration network to obtain a device parameter association diagram; the device parameter association diagram is a topological diagram that marks multiple parameter influence areas;

[0087] Input the device parameter association diagram into the analysis network to obtain multiple analysis dimensions.

[0088] It should be noted that the determination of the influence weight of the state parameters on each segment of the equipment current waveform data through the analysis network mentioned in the present invention is a key step in achieving accurate power equipment status assessment. The anomaly detection network is a model for identifying abnormal characteristics of equipment. It can extract the degree of deviation of key parameters from the equipment operating status data to form a first state feature set. The first state feature set is a quantitative representation of the deviation of the key parameters of the equipment from the normal range, reflecting the degree of abnormality of the equipment operating status. The position calibration network is used to generate an equipment parameter association diagram, which is a topological diagram that marks the areas where multiple parameters affect the equipment operating status. By inputting the equipment parameter association diagram into the analysis network, multiple analysis dimensions can be obtained. These analysis dimensions provide a multi-angle analysis basis for the subsequent equipment operating status assessment, thereby achieving comprehensive monitoring and precise control of the power equipment operating status.

[0089] Specifically, the input to the anomaly detection network is device operating status data, which includes device operating parameters such as load rate, operating duration, and startup frequency. The first state feature set is derived by analyzing the degree of deviation of these parameters. For example, the degree of deviation of the load rate can be determined by calculating the difference between the current load rate and the historical average load rate. The device parameter association graph is a topological structure that represents the associations between parameters in the form of nodes and edges. Nodes represent device operating parameters, and edges represent the mutual influence between parameters. The analysis network takes the device parameter association graph as input and, by analyzing the parameter associations in the graph, outputs multiple analysis dimensions. These analysis dimensions can represent different aspects of the impact of parameters on the device operating status, such as the degree of influence of the device startup frequency on the current waveform, the degree of influence of the load rate on the current waveform, and so on. Each analysis dimension corresponds to a specific parameter or parameter combination. Through these analysis dimensions, a more comprehensive understanding of the impact of the device operating status on the current waveform can be achieved.

[0090] Preferably, the anomaly detection network can be constructed using a machine learning algorithm, such as a support vector machine (SVM) or a convolutional neural network (CNN) within a deep learning algorithm. When constructing the anomaly detection network, a large amount of equipment operating status data must be collected as training samples. These samples include data from both normal and abnormal operating states. Through training, the model can learn the characteristic patterns of equipment operating status parameters under abnormal conditions, enabling accurate identification of equipment anomaly characteristics in practical applications. The position calibration network can be constructed using a graph neural network (GNN) algorithm, taking as input the feature vectors of the equipment operating status data and, through graph structure learning, outputting a device parameter association graph. When processing the device parameter association graph, feature extraction can be performed on each node in the graph, such as calculating the node's degree centrality or closeness centrality to determine its importance within the graph. The analysis network can be constructed using a multi-layer perceptron (MLP) or other deep learning model, taking as input the feature vectors of the device parameter association graph and outputting multiple analysis dimensions. When calculating the analysis dimensions, feature dimensionality reduction and cluster analysis can be performed on the equipment operating status data to extract the parameters or parameter combinations that have the greatest impact on the equipment operating status, thereby forming multiple analysis dimensions. These analysis dimensions can provide a more refined analytical basis for subsequent equipment operating status assessments.

[0091] In some embodiments, the training method of the analysis network includes:

[0092] Acquire multiple training data sets; the training data sets are historical device parameter association diagrams;

[0093] Obtaining a power state vector corresponding to the training data set as labeled data;

[0094] Inputting the training data set into the analysis network to obtain a training analysis dimension;

[0095] Based on the training analysis dimension and the labeled data, the training data set is divided to obtain a plurality of training partitioned data groups and corresponding core nodes; the training partitioned data groups are data subsets divided according to the training analysis dimension; the core nodes represent key positions of the first partitioned data group on the labeled reference curve;

[0096] Based on the training partition data group and the core node, the parameter influence degree is calculated with the main path as the benchmark to obtain multiple training partition load total values ​​and corresponding data intervals; the data interval represents the coverage range of the parameter influence area in the training partition data group; the training partition load total value represents the amplitude change of each parameter in the training partition data group during the data interval;

[0097] Calculating a parameter influence index based on a plurality of training partition load total values ​​and corresponding data intervals; wherein the plurality of training partition load total values ​​correspond to one parameter influence index; and wherein a plurality of training partition data groups correspond to obtaining a plurality of parameter influence indexes; wherein the parameter influence index represents the intensity of influence of the parameters of the entire training partition data group on the current waveform;

[0098] Dividing the multiple parameter influence indicators by the sum of the multiple parameter influence indicators to obtain an influence weight coefficient; the influence weight coefficient represents the proportion of the influence of the multiple training divided data groups on the current waveform;

[0099] A loss function is calculated based on the impact weight coefficient and the power state vector to optimize the analysis network.

[0100] It should be noted that the analysis network training method is a key step in achieving accurate analysis of device operating status. The training dataset is a historical device parameter correlation diagram that reflects the interrelationships between device operating parameters. The labeled data is the power state vector corresponding to the training dataset, which provides a target reference for the training process. By inputting the training dataset into the analysis network, the training analysis dimension is obtained, and the training dataset is then divided into training partitioned data groups and core nodes. Core nodes are key locations of the training partitioned data groups on the labeled reference curve and are used to determine the degree of parameter influence. By calculating parameter influence indicators and influence weight coefficients, the analysis network is ultimately optimized to enable more accurate analysis of device operating status.

[0101] Specifically, the training dataset consists of device parameter association graphs generated from historical device operating status data. These association graphs demonstrate the correlations between different parameters through topological structures. The labeled data is the power state vector, a multidimensional vector where each dimension corresponds to a characteristic of the device operating status, reflecting the state changes of the device under different operating conditions. The training analysis dimension is a feature dimension extracted by the analysis network from the training dataset to distinguish the factors influencing different device operating states. The partitioned data groups are data subsets divided according to the training analysis dimension. Each subset contains a specific parameter combination and its corresponding current waveform data. The core nodes are key locations of the partitioned data groups on the labeled reference curve and are used to determine the boundaries of the partitioned data groups. The parameter influence index is calculated by calculating the total training partition load and the data interval. It reflects the strength of the parameter's influence on the current waveform. The influence weight coefficient is obtained by normalizing the parameter influence index and measures the relative importance of different parameters on the current waveform. These parameters and indicators together form the basis for the analysis network training.

[0102] Preferably, the training of the analysis network can be achieved through the following steps: First, a large amount of historical equipment operating status data is collected, including the current waveform data, operating parameters, etc. of the equipment, and an equipment parameter association diagram is generated as a training data set. Secondly, according to the actual operating status of the equipment, a corresponding power state vector is generated as the annotation data. The training data set is input into the analysis network, and features are extracted through a neural network algorithm, such as a convolutional neural network or a recurrent neural network, to obtain a training analysis dimension. Based on the training analysis dimension, the training data set is divided to obtain multiple training partitioned data groups, and the core nodes of each partitioned data group are determined by a clustering algorithm. The parameter influence index is obtained by calculating the total value of the training partition load and the corresponding data interval of each partitioned data group. The parameter influence index is normalized to obtain the influence weight coefficient. Finally, the loss function is calculated based on the influence weight coefficient and the power state vector, and the weight and bias parameters of the analysis network are optimized through the back propagation algorithm, so that the network can more accurately analyze the equipment operating status.

[0103] In some embodiments, generating a power state vector based on the device operating state data and the device current waveform data includes:

[0104] Using the correlation curve between the device operating status data and the device current waveform data as a reference curve;

[0105] Segmenting the device current waveform data along a direction perpendicular to the reference curve to obtain a plurality of first segmented data sets;

[0106] Accumulating the amplitude parameters in the first segmented data set to obtain a segment load value; obtaining multiple segment load values ​​corresponding to multiple first segmented data sets;

[0107] The plurality of segment load values ​​are arranged in sequence according to the time sequence positions of the plurality of first segment data sets in the device current waveform data to generate a power state vector.

[0108] It should be noted that the process of generating the power state vector mentioned in this invention combines device operating status data with device current waveform data to quantify the impact of the device operating status on the current waveform. This process segments the device current waveform data along the reference curve and calculates the load value of each segment, ultimately generating a power state vector that reflects the device operating status. The power state vector is a multidimensional data structure that represents the degree to which changes in device operating status parameters affect the current waveform amplitude and serves as the basis for subsequent device status assessment and control.

[0109] Specifically, the correlation curve between the equipment operating status data and the equipment current waveform data is used as a reference curve and is established based on the relationship between the equipment operating parameters and the current waveform data. The equipment current waveform data is segmented in a direction perpendicular to the reference curve, that is, the current waveform data is divided into multiple time intervals or amplitude intervals, and the current amplitude of each interval is accumulated to obtain the segment load value. These segment load values ​​are arranged in sequence according to their time sequence positions in the equipment current waveform data to form a power state vector. The segment load value reflects the total change in the equipment current within a specific time interval, and the power state vector integrates the overall impact of the equipment operating status parameters on the current waveform amplitude. For example, if the load of the equipment increases within a certain time period, the amplitude of the current waveform will also increase accordingly. This change will be recorded in the corresponding segment load value and reflected in the power state vector.

[0110] Preferably, the process of generating the power state vector can be refined by the following steps: First, according to the characteristics of the equipment operating status data and the equipment current waveform data, select appropriate time intervals or amplitude intervals to segment the current waveform data. For example, a complete current cycle can be divided into several time periods of equal length, or different intervals can be divided according to the range of change of the current amplitude. Secondly, for each segmented interval, the cumulative sum of the current amplitudes in the interval is calculated to obtain the segment load value. This calculation process can be achieved through numerical integration or simple accumulation operations, depending on the sampling method and accuracy of the current waveform data. Finally, all segment load values ​​are arranged in their time sequence in the current waveform to form a power state vector. In practical applications, the number and method of segmentation can be adjusted according to the operating characteristics and monitoring requirements of the equipment to optimize the generation process of the power state vector, thereby more accurately reflecting the impact of the equipment operating status on the current waveform.

[0111] In some embodiments, the training data set is divided based on the training analysis dimension and the labeled data to obtain multiple training divided data groups and corresponding core nodes, including:

[0112] Using the first segmented data set of the device current waveform data corresponding to the labeled data as a labeled first segmented set;

[0113] Using the reference curve corresponding to the first segment set as the labeled reference curve;

[0114] Marking the first segmented set according to the positions of the labeled reference curve in the training data set to obtain a plurality of first labeled positions;

[0115] The midpoint of two adjacent first-marked positions is taken as the core node;

[0116] Determine the angle difference from the annotated reference curve as the training analysis dimension and the dividing boundary passing through the core node;

[0117] The training data set is divided using the division boundary to obtain a training division data set.

[0118] It should be noted that the process of dividing the training data set mentioned in the present invention is implemented based on the device current waveform data and the reference curve corresponding to the labeled data. This process determines the position of the core node by marking and dividing the data set, thereby providing a basis for subsequent analysis. The labeled first segment set is a specific data set divided according to the device current waveform data corresponding to the labeled data, and the labeled reference curve is a reference curve generated based on these data sets, which is used to guide the division of the data set. The core node is the key position for dividing the data set, and the dividing boundary is determined based on the angle difference between the core node and the labeled reference curve, which is used to divide the data set into multiple subsets. In this way, the impact of the equipment operating status data on the current waveform can be analyzed more accurately.

[0119] Specifically, the annotated first segment set refers to a plurality of data subsets divided according to the device current waveform data corresponding to the annotated data, and these subsets reflect the current characteristics of the device under different operating conditions. The annotated reference curve is generated based on the annotated first segment set. It is a standard curve used to describe the typical characteristics of the device current waveform. The first marking position is obtained by marking the annotated first segment set according to the position of the annotated reference curve in the training data set. These positions are used to determine the reference points for dividing the data set. The core node is the midpoint of two adjacent first marking positions, which represents the key position of the divided data group on the annotated reference curve. The dividing boundary is determined based on the angle difference from the annotated reference curve, and is used to divide the training data set into multiple subsets. These subsets can be used to further analyze the impact of the device operating state on the current waveform.

[0120] Preferably, the process of dividing the training data set can be refined by the following steps: First, based on the device current waveform data corresponding to the labeled data, the current waveform is divided into multiple representative segments to form a labeled first segment set. Then, a labeled reference curve is generated based on these segments. This curve can be an average curve of the device current waveform or other statistical curve, which is used to describe the typical characteristics of the device current waveform. Next, the position of the labeled first segment set in the training data set is marked to obtain multiple first marked positions. The position of the core node is determined by calculating the midpoint of two adjacent first marked positions. Finally, based on the angle difference of the labeled reference curve and the position of the core node, the division boundary is determined to divide the training data set into multiple subsets. In practical applications, the number of segments and the angle difference of the division boundary can be adjusted according to the characteristics of the device current waveform and the complexity of the operating status data to optimize the data set division process, thereby more accurately analyzing the impact of the device operating status on the current waveform.

[0121] In some embodiments, the parameter influence degree is calculated based on the training partition data group and the core node with the main path as the benchmark to obtain multiple training partition load total values ​​and corresponding data intervals, including:

[0122] Determining a main path of the training partition data group according to the training partition data group and the core node;

[0123] Based on the training partition data set and the main path, the parameter influence is mapped to the main path to obtain a plurality of training partition load total values; each position on the main path corresponds to a training partition load total value;

[0124] Get the coordinates of the positions on the main path in the training data set as the division positions;

[0125] According to the coordinates of the partitioned position and the core node, the distance value is calculated using the Euclidean distance formula to obtain the data interval;

[0126] The training partition load total value is associated with the corresponding data interval and stored.

[0127] It should be noted that the process of calculating the degree of parameter influence based on the training partition data group and the core node mentioned in the present invention is to analyze the main path of the training partition data group and map the parameter influence to the main path, thereby obtaining multiple training partition load total values ​​and corresponding data intervals. The core of this process is to quantify the degree of influence of the parameters on the current waveform through main path analysis, and further clarify the scope and distribution of such influence through the calculation of data intervals. The main path is the main path of parameter influence in the partition data group, the training partition load total value reflects the intensity of the parameter influence on the main path, and the data interval represents the coverage of the parameter influence area in the partition data group.

[0128] Specifically, the training partition data group is a data subset divided according to the analysis dimension, which contains the equipment operating status data and the corresponding current waveform data. The core node is the key position of the partition data group on the annotated reference curve, which is used to determine the starting point and end point of the main path. The main path is the path connecting the core node and the key position in the partition data group, which reflects the main direction of parameter influence. The total value of the training partition load is obtained by accumulating the parameter influence on the main path, which represents the total influence intensity of the parameter on the main path. The partition position is the coordinate of the position on the main path in the training data set, which is used to calculate the data interval. The data interval is obtained by calculating the Euclidean distance between the partition position and the core node coordinates, which reflects the coverage of the parameter influence area. By associating and storing the total value of the training partition load with the data interval, data support can be provided for subsequent parameter impact analysis.

[0129] Preferably, the process of calculating the degree of parameter influence can be refined by the following steps: First, the main path is determined based on the training partition data group and the core node. If the data group is of regular form, a reference line can be generated by connecting specific key feature points and the main path is determined. If the data group is of irregular form, the main path needs to be determined by feature identification points and marking feature points. Then, a plurality of discrete partition reference positions are selected on the main path, and these positions are evenly distributed along the main path. By determining a partition reference plane perpendicular to the main path, the amplitude parameters on the partition reference plane are accumulated in the partition data group to obtain the total value of the training partition load. Then, the position coordinates on the main path are obtained as the partition position, and the distance between the partition position and the core node is calculated by the Euclidean distance formula to obtain the data interval. Finally, the total value of the training partition load is associated with the corresponding data interval and stored for subsequent analysis. In practical applications, the number and distribution of partition reference positions can be adjusted according to the form and complexity of the data group to optimize the calculation process of the parameter influence degree, thereby more accurately quantifying the influence of the parameters on the current waveform.

[0130] In some embodiments, mapping the parameter influence onto the main path based on the training partition data set and the main path to obtain a plurality of training partition load total values ​​includes:

[0131] Selecting discrete positions on the main path as partition reference positions; multiple partition reference positions are evenly distributed along the main path;

[0132] Determine a plane passing through the division reference position and perpendicular to the main path as the division reference plane;

[0133] In the training partition data group, the amplitude parameters on the partition reference plane are accumulated to obtain the total training partition load value; multiple partition reference planes correspond to multiple training partition load total values.

[0134] It should be noted that the process of calculating the total value of the training divided load by dividing the reference plane mentioned in the present invention is based on discrete positions on the main path. The dividing reference plane is a plane perpendicular to the main path. By accumulating the amplitude parameters on these planes, the total value of the training divided load can be obtained. This method can effectively quantify the intensity of the parameter influence at different positions on the main path, thereby providing more accurate data support for analyzing the impact of the equipment operating status on the current waveform. By performing calculations on multiple divided reference planes, the distribution of the influence of the parameters on the main path can be comprehensively evaluated.

[0135] Specifically, the partition reference positions are discrete positions on the main path, which are evenly distributed along the main path and are used to determine the position of the partition reference plane. The partition reference plane is a plane that passes through the partition reference position and is perpendicular to the main path, and is used to intercept the amplitude parameters in the training partition data group. The amplitude parameter refers to the amplitude of the current waveform data on the partition reference plane, which reflects the current intensity of the device at that position. By accumulating the amplitude parameters on each partition reference plane, the total training partition load value can be obtained. These total values ​​represent the parameter influence intensity at different positions on the main path. Multiple partition reference planes correspond to multiple training partition load total values, and these total values ​​can be used to further analyze the impact of the operating status of the equipment on the current waveform. For example, if the amplitude parameter on a certain partition reference plane is large, it means that the current intensity of the equipment at that position is high, which may be related to the operating status of the equipment.

[0136] Preferably, the process of calculating the total value of the training partition load can be refined by the following steps: First, the number and distribution of the partition reference positions are determined according to the length of the main path and the accuracy required for analysis. The number of partition reference positions can be adjusted according to the complexity of the device current waveform and the size of the data set. Then, for each partition reference position, a partition reference plane passing through the position and perpendicular to the main path is determined. On the partition reference plane, the amplitude parameters in the training partition data group are extracted, and these parameters are accumulated to obtain the total value of the training partition load at that position. In this way, the total value of the training partition load can be calculated at multiple positions on the main path, thereby forming a distribution diagram of the parameter influence intensity. In practical applications, the number and distribution of the partition reference positions can be adjusted according to the operating characteristics of the equipment and the characteristics of the current waveform to optimize the calculation process and improve the accuracy and efficiency of the analysis.

[0137] In some embodiments, the calculating of the parameter impact index based on the total values ​​of the multiple training partition loads and the corresponding data intervals includes:

[0138] Sum multiple data intervals to get the interval sum;

[0139] Divide the individual data interval by the sum of the intervals to obtain the interval influence coefficient;

[0140] Multiplying the total value of the training division load by the corresponding interval influence coefficient to obtain a division influence value;

[0141] All partition influence values ​​are summed to obtain the parameter influence index; multiple training partition data groups correspond to multiple parameter influence indexes.

[0142] It should be noted that the process of calculating the parameter impact index mentioned in the present invention is based on the total value of the training partition load and the corresponding data interval. This process provides a comprehensive evaluation index for subsequent analysis by quantifying the degree of influence of each training partition data group on the current waveform. The parameter impact index reflects the intensity of the influence of the parameters of the entire training partition data group on the current waveform, and is obtained by combining the total value of the training partition load with the relative importance of the data interval. In this way, the influence of the equipment operating status parameters on the current waveform can be more comprehensively evaluated, providing important data support for the status evaluation and control of the power system.

[0143] Specifically, the data interval refers to the coverage of the parameter influence area in the training partition data group, which reflects the spatial distribution characteristics of the partition data group on the main path. The interval sum is the cumulative value of all data intervals, which is used to normalize the influence of a single data interval. The interval influence coefficient is the ratio of a single data interval to the interval sum, which indicates the relative importance of each data interval in the overall distribution. The total value of the training partition load is the cumulative value of the amplitude parameter on the partition reference plane, which reflects the influence intensity of the parameter at a specific position on the main path. The partition influence value is the product of the total value of the training partition load and the corresponding interval influence coefficient. It combines the parameter influence intensity and the relative importance of the data interval. The parameter influence index is the cumulative sum of all partition influence values. It is a comprehensive index used to quantify the degree of influence of the entire training partition data group on the current waveform. In this way, the influence of the equipment operating status parameters on the current waveform can be more accurately evaluated, providing a basis for subsequent status evaluation.

[0144] Preferably, the process of calculating the parameter impact index can be refined by the following steps: First, the sum of all data intervals, i.e., the interval sum, is calculated. Then, for each data interval, the ratio of the interval sum to the interval sum is calculated to obtain the interval impact coefficient. Next, the total load of each training partition is multiplied by its corresponding interval impact coefficient to obtain the partition impact value. Finally, all partition impact values ​​are accumulated to obtain the parameter impact index.

[0145] Furthermore, in practical applications, the data interval division method and the calculation accuracy of the total training load can be adjusted according to the operating characteristics of the equipment and the characteristics of the current waveform to optimize the calculation process of the parameter impact index. For example, if the current waveform of the equipment changes drastically, the number of data intervals can be increased to more accurately evaluate the impact of the parameters. In this way, the impact of the equipment operating status parameters on the current waveform can be more accurately reflected, providing more reliable data support for power system status assessment and control.

[0146] In some embodiments, determining the main path of the training partitioned data group according to the training partitioned data group and the core node includes:

[0147] If the training divided data group is in a regular form, obtaining key feature points;

[0148] Connect two specific key feature points to generate a reference line;

[0149] Determine key locations on the reference line;

[0150] Associate core nodes with key locations to determine the main path.

[0151] It should be noted that the process of determining the main path of the training partition data group mentioned in the present invention is a processing method for regular morphological data groups. The main path is used to describe the main direction and path of parameter influence, which is determined by connecting key feature points. Key feature points are positions with significant features in the data group, such as peak points or trough points of the current waveform. The reference line is a straight line formed by connecting two specific key feature points, which is used to determine the basic direction of the main path. The key position is a specific point on the reference line, which is used to further determine the precise position of the main path. By associating the core node with the key position, the main path can be determined, thereby providing a basis for subsequent parameter impact analysis.

[0152] Specifically, the training partition data group refers to the data subset divided according to the analysis dimension, and these data subsets contain the equipment operation status data and the corresponding current waveform data. Regular morphology means that the shape of the data group is relatively regular, such as linear or approximately linear distribution. Key feature points are positions with significant features in the data group, such as peak points, trough points or other significant change points of the current waveform. The reference line is a straight line formed by connecting two specific key feature points, which is used to determine the basic direction of the main path. The key position is a specific point on the reference line, such as the midpoint or a position point of a specific proportion, which is used to further determine the precise position of the main path. The core node is the key position of the partition data group on the annotated reference curve, which is used to determine the starting point or end point of the main path. By associating the core node with the key position, the main path can be determined, thereby providing a clear direction for subsequent parameter impact analysis.

[0153] Preferably, the process of determining the main path can be refined through the following steps: First, identify key feature points in the training partitioned data set. These points are typically peaks, troughs, or other significant change points in the current waveform. Then, select two specific key feature points, such as the starting and ending points of the current waveform, and connect them to form a reference line. Next, determine a key location on the reference line, such as the midpoint or a location at a specific ratio. Finally, associate the core node with the key location to form the main path.

[0154] Furthermore, in practical applications, appropriate key feature points and key locations can be selected based on the data set's morphology and the characteristics of the current waveform. For example, if the current waveform is relatively flat, the waveform's starting and ending points can be selected as key feature points. If the current waveform has multiple peaks or troughs, the two most significant points can be selected as key feature points. This approach allows for more accurate identification of the main path, providing a more reliable foundation for subsequent parameter impact analysis.

[0155] In some embodiments, determining the main path of the training partitioned data group according to the training partitioned data group and the core node includes:

[0156] If the training data group is irregular, obtain the feature identification points;

[0157] The feature identification points containing the core nodes are used as marking feature points;

[0158] The associated area of ​​the marked feature points is used as the reference area;

[0159] Identify key locations in the reference area;

[0160] Connect key locations and marked feature points to determine the main path.

[0161] It should be noted that the process of determining the main path of the training partition data group mentioned in the present invention is a processing method for irregular morphology data groups. Irregular morphology data groups mean that their shape and distribution do not conform to simple linear or other regular patterns, so different methods are needed to determine the main path. Feature identification points are points in the data group used to identify features, and marked feature points are feature identification points containing core nodes, which are used to determine the starting point or key position of the main path. The reference area is the associated area of ​​the marked feature points, which is used to further determine the direction and range of the main path. By connecting the key positions and the marked feature points, the main path can be determined, thereby providing a clear direction for the subsequent parameter impact analysis.

[0162] Specifically, irregular shape means that the shape and distribution of the data group are more complex and do not conform to simple linear or other regular patterns. Feature identification points are points in the data group used to identify features, such as local peak points, trough points or other significant change points of the current waveform. Marked feature points are feature identification points that contain core nodes, which are used to determine the starting point or key position of the main path. The reference area is the associated area of ​​the marked feature point, which usually refers to the area within a certain range centered on the marked feature point, and is used to determine the direction and range of the main path. The key position is a specific point in the reference area, such as the geometric center point or other significant position point of the reference area. By connecting the key position with the marked feature point, the main path can be determined, thereby providing a clear direction for subsequent parameter impact analysis. The core node is the key position of the data group on the marked reference curve, which is used to determine the starting point or end point of the main path.

[0163] Preferably, the process of determining the main path can be refined by the following steps: First, identify the characteristic identification points in the training partition data group, which are usually local peak points, trough points or other significant change points of the current waveform. Then, select the characteristic identification points containing the core nodes as the marking feature points, which are used to determine the starting point or key position of the main path. Next, determine the associated area of ​​the marking feature points, such as the area within a certain range centered on the marking feature points. Within the reference area, determine a key position, such as the geometric center point of the reference area or other significant position points. Finally, connect the key position and the marking feature points to form the main path.

[0164] Furthermore, in practical applications, appropriate feature identification points and key locations can be selected based on the data set's morphology and the characteristics of the current waveform. For example, if the current waveform has multiple local peaks or troughs, the most significant points can be selected as feature identification points. This approach allows for more accurate identification of the primary path, providing a more reliable foundation for subsequent parameter impact analysis.

[0165] The aforementioned embodiments of the present invention have the following beneficial effects: This method can implement automated load classification and status monitoring of power equipment based on power operation data. By analyzing the correlation between current waveforms and operating parameters, a power state vector is generated, accurately reflecting the operating characteristics of the equipment. Through multi-dimensional parameter weight analysis and data partitioning, abnormal equipment conditions can be accurately identified and control instructions generated, significantly improving the intelligence level and response efficiency of remote monitoring of power systems.

[0166] This invention optimizes the power data analysis process. By constructing a parameter association diagram and locating core nodes, it can accurately calculate the weight of each state parameter's impact on the current waveform. Using segmented load calculation and main path mapping methods, it can quantitatively assess the operating status of equipment. By training the network to optimize the analysis dimension division, it can improve the accuracy of anomaly detection and the system's adaptability, providing reliable data support for power equipment operation and maintenance.

[0167] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, server, mobile phone, or tablet.

[0168] The above descriptions merely illustrate some preferred embodiments of the present invention and the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for remotely controlling power data, characterized in that: Applied to remote control system, which includes data acquisition module and processor: The data acquisition module is used to obtain power operation data; the power operation data includes voltage value, current value and power factor; The processor is configured to perform the following steps: Inputting the power operation data into a power load analysis network to obtain a plurality of load classification results; the load classification results indicate the load types corresponding to the power equipment; According to the load classification result, obtaining device current waveform data and corresponding device operating status data; The device current waveform data represents a periodic current variation curve generated when the power device is running; The device operating status data represents a set of device operating parameters corresponding to when recording device current waveform data; generating a power state vector based on the device operating state data and the device current waveform data; wherein the power state vector represents the effect of parameter changes in the device operating state data on the amplitude of the current waveform; By analyzing the network, based on the device operating status data, determining the influence weight of the status parameter on each section of the device current waveform data, and obtaining multiple analysis dimensions; Based on the analysis dimension and the device operating status data, the device operating status data is divided into different dimensions to obtain a plurality of divided data groups; the divided data groups are sub-data sets including the device current waveform data; Through the status assessment network, based on the divided data group, it is determined whether the equipment operating status is abnormal and a control instruction is generated; By analyzing the network, based on the device operating status data, the influence weight of the status parameter on each segment in the device current waveform data is determined to obtain multiple analysis dimensions, including: Inputting the equipment operation status data into an anomaly detection network to identify equipment anomaly characteristics and obtain a first state feature set; the first state feature set represents the degree of deviation of key parameters of the equipment; Inputting the first state feature set into a position calibration network to obtain a device parameter association diagram; the device parameter association diagram is a topological diagram that marks multiple parameter influence areas; Input the equipment parameter association diagram into the analysis network to obtain multiple analysis dimensions; The training method of the analysis network includes: Acquire multiple training data sets; the training data sets are historical device parameter association diagrams; Obtaining a power state vector corresponding to the training data set as labeled data; Inputting the training data set into the analysis network to obtain a training analysis dimension; Based on the training analysis dimension and the labeled data, the training data set is divided to obtain a plurality of training partitioned data groups and corresponding core nodes; the training partitioned data groups are data subsets divided according to the training analysis dimension; the core nodes represent key positions of the first partitioned data group on the labeled reference curve; Based on the training partition data group and the core node, the parameter influence degree is calculated with the main path as the benchmark to obtain multiple training partition load total values ​​and corresponding data intervals; the data interval represents the coverage range of the parameter influence area in the training partition data group; the training partition load total value represents the amplitude change of each parameter in the training partition data group during the data interval; Calculating a parameter influence index based on a plurality of training partition load total values ​​and corresponding data intervals; wherein the plurality of training partition load total values ​​correspond to one parameter influence index; and wherein a plurality of training partition data groups correspond to obtaining a plurality of parameter influence indexes; wherein the parameter influence index represents the intensity of influence of the parameters of the entire training partition data group on the current waveform; Dividing the multiple parameter influence indicators by the sum of the multiple parameter influence indicators to obtain an influence weight coefficient; the influence weight coefficient represents the proportion of the influence of the multiple training divided data groups on the current waveform; A loss function is calculated based on the impact weight coefficient and the power state vector to optimize the analysis network.

2. The method for remotely controlling power data according to claim 1, characterized in that: The generating of the power state vector based on the device operating state data and the device current waveform data includes: Using the correlation curve between the device operating status data and the device current waveform data as a reference curve; Segmenting the device current waveform data along a direction perpendicular to the reference curve to obtain a plurality of first segmented data sets; Accumulating the amplitude parameters in the first segmented data set to obtain a segment load value; obtaining multiple segment load values ​​corresponding to multiple first segmented data sets; The plurality of segment load values ​​are arranged in sequence according to the time sequence positions of the plurality of first segment data sets in the device current waveform data to generate a power state vector.

3. The method for remotely controlling power data according to claim 1, characterized in that: The training data set is divided based on the training analysis dimension and the labeled data to obtain multiple training divided data groups and corresponding core nodes, including: Using the first segmented data set of the device current waveform data corresponding to the labeled data as a labeled first segmented set; Using the reference curve corresponding to the first segment set as the labeled reference curve; Marking the first segmented set according to the positions of the labeled reference curve in the training data set to obtain a plurality of first labeled positions; The midpoint of two adjacent first-marked positions is taken as the core node; Determine the angle difference from the annotated reference curve as the training analysis dimension and the dividing boundary passing through the core node; The training data set is divided using the division boundary to obtain a training division data set.

4. The method for remotely controlling power data according to claim 1, characterized in that: The method of calculating the parameter influence based on the training partition data group and the core node and taking the main path as a benchmark to obtain a plurality of training partition load total values ​​and corresponding data intervals includes: Determining a main path of the training partition data group according to the training partition data group and the core node; Based on the training partition data set and the main path, the parameter influence is mapped to the main path to obtain a plurality of training partition load total values; each position on the main path corresponds to a training partition load total value; Get the coordinates of the positions on the main path in the training data set as the division positions; According to the coordinates of the partitioned position and the core node, the distance value is calculated using the Euclidean distance formula to obtain the data interval; The training partition load total value is associated with the corresponding data interval and stored.

5. The method for remotely controlling power data according to claim 4, characterized in that: The parameter influence is mapped onto the main path based on the training partition data group and the main path to obtain a plurality of training partition load total values, including: Selecting discrete positions on the main path as partition reference positions; multiple partition reference positions are evenly distributed along the main path; Determine a plane passing through the division reference position and perpendicular to the main path as the division reference plane; In the training partition data group, the amplitude parameters on the partition reference plane are accumulated to obtain the total training partition load value; multiple partition reference planes correspond to multiple training partition load total values.

6. The method for remotely controlling power data according to claim 1, characterized in that: The calculating of the parameter impact index based on the total values ​​of the multiple training division loads and the corresponding data intervals includes: Sum multiple data intervals to get the interval sum; Divide the individual data interval by the sum of the intervals to obtain the interval influence coefficient; Multiplying the total value of the training division load by the corresponding interval influence coefficient to obtain a division influence value; All the partition influence values ​​are summed up to obtain the parameter influence index; multiple training partition data groups correspond to multiple parameter influence indexes.

7. The method for remotely controlling power data according to claim 4, characterized in that: The determining of the main path of the training divided data group according to the training divided data group and the core node includes: If the training divided data group is in a regular form, obtaining key feature points; Connect two specific key feature points to generate a reference line; Determine key locations on the reference line; Associate core nodes with key locations to determine the main path.

8. The method for remotely controlling power data according to claim 4, characterized in that: The determining of the main path of the training divided data group according to the training divided data group and the core node includes: If the training data group is irregular, obtain the feature identification points; The feature identification points containing the core nodes are used as marking feature points; The associated area of ​​the marked feature points is used as the reference area; Identify key locations in the reference area; Connect key locations and marked feature points to determine the main path.

Images