An automatic measurement and control intelligent box transformer control method and system

Abstract

The application discloses an automatic measurement and control intelligent box transformer control method and system, relates to the technical field of smart grid control, and comprises the following steps: collecting multi-source data by using intelligent sensors, simulating a group of virtual neurons for each intelligent sensor, calculating activation values, summing the activation values, generating aggregated features, calculating the normalized mutual information between the aggregated features, calculating the convolution feature vector by using Chebyshev polynomials, and generating a dimension reduction feature matrix; calculating the speed of a window by using a particle interaction model, calculating the abnormal score of the window, calculating the KL divergence of adjacent windows by using a KL divergence formula, defining a target function by using a weighted sum, and generating a control sequence. The simulation of intelligent sensors and virtual neurons, the Chebyshev polynomial method and the dimension reduction feature matrix generation technology are introduced, the precision and efficiency of feature extraction are improved, the particle interaction model and the KL divergence analysis are adopted, and the accuracy of the control sequence is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid control, and in particular to an automatic measurement and control method and system for a smart box transformer. BACKGROUND

[0002] With the development of intelligent control technology, automatic intelligent transformer control methods have become an important research direction in power systems to improve work efficiency, reduce operating costs and improve safety. As an important part of the power network, smart substations (referred to as "smart box transformers") achieve remote monitoring and automatic control of power equipment by integrating advanced sensors, control algorithms and data analysis methods. With the rapid development of technologies such as the Internet of Things, big data analysis and machine learning, the data collection and analysis capabilities of smart sensors have improved significantly, thereby promoting the widespread application of smart box transformers.

[0003] Existing smart box transformer control methods still have deficiencies. Existing technologies have certain limitations in multi-source data fusion and intelligent decision-making. Existing smart sensor data analysis methods often rely on simple statistical analysis and lack deep mining of data dynamic changes and time series characteristics, resulting in systems that cannot fully utilize complex data for intelligent prediction and decision optimization. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an automatic measurement and control method and system for a smart box transformer, which solves the problem of certain limitations in multi-source data fusion and intelligent decision-making in existing technologies. Existing smart sensor data analysis methods often rely on simple statistical analysis and lack deep mining of data dynamic changes and time series characteristics, resulting in systems that cannot fully utilize complex data for intelligent prediction and decision optimization.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an automatic measurement and control method for a smart box transformer, comprising,

[0008] Collecting multi-source data using smart sensors, for each smart sensor, simulating a group of virtual neurons, calculating activation values, summing the activation values, generating aggregated features, calculating normalized mutual information between the aggregated features, using Chebyshev polynomials to calculate convolution feature vectors, and generating a reduced dimension feature matrix;

[0009] Set the window size using the square root rule, for each window, define the position vector, calculate the velocity of the window using the particle interaction model, calculate the anomaly score of the window, calculate the KL divergence of adjacent windows using the KL divergence formula, define the objective function using the weighted sum, and generate the control sequence;

[0010] The control sequence generates control instructions and executes, collects feedback data for monitoring and optimization;

[0011] The collected and analyzed multi-source data are stored.

[0012] As a preferred scheme of the intelligent box transformer control method of the automatic measurement and control of the application, wherein: the multi-source data collected by the intelligent sensor, for each intelligent sensor, simulate a group of virtual neurons, calculate the activation value, sum the activation value, generate the aggregated feature, calculate the normalized mutual information between the aggregated features, calculate the convolution feature vector using the Chebyshev polynomial, and generate the dimension reduction feature matrix, including:

[0013] The multi-source data collected by the intelligent sensor includes pressure, current, voltage and temperature data;

[0014] The multi-source data is sorted in time sequence to generate a time series and is standardized;

[0015] For each intelligent sensor, the sample entropy of the standardized time series is calculated using the standard sample entropy algorithm, and is sorted in ascending order, the sample entropy is calculated in all sample entropies using the quantile mapping method, the normalized quantile is obtained, the normalized quantile is linearly mapped to the neuron number range, and the neuron number of the intelligent sensor is calculated;

[0016] The autocorrelation function value of the intelligent sensor is calculated using the autocorrelation function formula, the first local maximum value in the autocorrelation function value is identified as the main period using peak detection, and the time constant is calculated using linear scaling adjustment;

[0017] For each intelligent sensor, a group of virtual neurons is simulated, the activation value is calculated, and the activation value is summed to generate an aggregated feature;

[0018] The intelligent sensor is defined as a node, the connection between the nodes is defined as an edge, an undirected graph is constructed, the normalized mutual information between the aggregated features is calculated, and the edge weight is defined;

[0019] The split threshold is set using the percentile method, the edge weight greater than the split threshold is filtered, the adjacency matrix of the graph is constructed, the sum of the corresponding rows of the node in the adjacency matrix is calculated to obtain the degree of the node, the normalized Laplacian matrix is calculated, the normalized Laplacian matrix is linearly transformed, the scaling Laplacian matrix is calculated, and the Chebyshev polynomial matrix is recursively calculated based on the scaling Laplacian matrix.

[0020] The convolution feature vector is calculated using Chebyshev polynomials, and the convolution feature vector is assembled into a convolution feature matrix using matrix assembly;

[0021] The covariance matrix of the convolution feature matrix is calculated using the sample covariance matrix method, the eigenvalues and eigenvectors of the covariance matrix are calculated using random singular value decomposition, the eigenvectors corresponding to the first o largest eigenvalues are selected to form a projection matrix, and the convolution feature matrix is projected into a low-dimensional space using the projection matrix to generate a dimensionality reduction feature matrix and perform normalization processing.

[0022] As a preferred scheme of the intelligent box transformer control method of the automatic measurement and control, wherein: the square root rule is used to set the window size, for each window, the position vector is defined, the particle interaction model is used to calculate the speed of the window, and the abnormal score of the window is calculated, including:

[0023] The square root rule is used to set the window size, the normalized dimensionality reduction feature matrix is segmented according to the window size, the time series slice is used to extract the subsequence, and the mean and standard deviation of the subsequence are calculated as the statistical characteristics of the window, for each window, the position vector is defined ;

[0024] Based on the position vector, the particle interaction model is used to calculate the speed of the window, and the position vector is updated using the particle swarm optimization;

[0025] The mean of the updated position vector is calculated and defined as the group center, the mean absolute deviation is used to calculate the abnormal score of the window, and the abnormal score is normalized.

[0026] As a preferred scheme of the intelligent box transformer control method of the automatic measurement and control, wherein: the KL divergence formula is used to calculate the KL divergence of adjacent windows, and the weighted sum is used to define the target function to generate the control sequence, including:

[0027] The normalized abnormal score is defined as the mean of the Gaussian distribution, the empirical standard deviation is used to calculate the standard deviation, and the KL divergence formula is used to calculate the KL divergence of adjacent windows;

[0028] The current state parameters are obtained using the standard industrial protocol, the current state parameters are defined as the control parameter vector using the vector construction, and the control parameter vector is assigned an initial value using the default value;

[0029] The weighted sum is used to define the target function, the central difference method is used to calculate the numerical gradient of the control parameter vector, the control parameter vector is updated based on the numerical gradient using the gradient descent method, and the final control vector is obtained, each final control vector is connected in the order of the time window to generate the control sequence.

[0030] As a preferred scheme of the intelligent box transformer control method of automatic measurement and control, wherein: the control sequence generates control instructions and executes, including:

[0031] The control sequence is converted into control instructions using a control instruction mapping method;

[0032] The generated control instructions are transmitted and executed through the communication interface Modbus.

[0033] As a preferred scheme of the intelligent box transformer control method of automatic measurement and control, wherein: the feedback data is collected for monitoring and optimization, including:

[0034] The feedback data is sequentially connected according to the time window to generate an actual state sequence;

[0035] The absolute error value method is used to calculate the error between the control sequence and the actual state sequence;

[0036] The statistical analysis method is used to set a judgment threshold, the error is compared with the judgment threshold, the error rate greater than or equal to the judgment threshold is adjusted using the PID control algorithm, and the adjustment is stopped when the difference is less than the judgment threshold, and the feedback data is continuously monitored.

[0037] As a preferred scheme of the intelligent box transformer control method of automatic measurement and control, wherein: the collected and analyzed multi-source data is stored, including:

[0038] The collected multi-source data and the analyzed control sequence are stored in the central database, and security access measures are set, the central database stores the data in the cloud backup, and the integrity of the stored data and the backup data is detected regularly, and the integrity detection record is generated after the detection is completed and is stored in the central database.

[0039] Secondly, the application provides an automatic measurement and control intelligent box transformer control system, including,

[0040] The collection dimension reduction module is used to collect multi-source data using intelligent sensors, simulate a group of virtual neurons for each intelligent sensor, calculate the activation value, sum the activation value, generate aggregated features, calculate the normalized mutual information between the aggregated features, calculate the convolution feature vector using the Chebyshev polynomial, and generate the dimension reduction feature matrix;

[0041] The scoring control module is used to set the window size using the square root rule, define the position vector for each window, calculate the speed of the window using the particle interaction model, calculate the abnormal score of the window, calculate the KL divergence of adjacent windows using the KL divergence formula, define the target function using the weighted sum, and generate the control sequence;

[0042] An execution monitoring module is configured to generate control instructions and execute the control instructions, and collect feedback data for monitoring and optimization.

[0043] A storage module is configured to store the collected and analyzed multi-source data.

[0044] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the automatic measurement and control intelligent box transformer control method according to the first aspect of the present application.

[0045] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any step of the automatic measurement and control intelligent box transformer control method according to the first aspect of the present application.

[0046] The present application has the following beneficial effects: the present application uses intelligent sensors to collect multi-source data, simulates a group of virtual neurons for each intelligent sensor, calculates an activation value, sums the activation values, generates aggregated features, calculates normalized mutual information between the aggregated features, uses Chebyshev polynomials to calculate a convolution feature vector, and generates a dimension-reduced feature matrix; uses a square root rule to set a window size, defines a position vector for each window, uses a particle interaction model to calculate the speed of the window, calculates an anomaly score of the window, uses a KL divergence formula to calculate the KL divergence of adjacent windows, uses a weighted sum to define an objective function, and generates a control sequence; the accuracy and efficiency of feature extraction are improved, the relevance analysis between data is enhanced, and the accuracy of the control sequence is improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0048] Fig. 1 A flowchart of the automatic measurement and control intelligent box transformer control method in embodiment 1.

[0049] Fig. 2 A schematic diagram of the automatic measurement and control intelligent box transformer control system in embodiment 1.

[0050] Fig. 3 A schematic diagram of generating control instructions and executing in embodiment 1.

[0051] Fig. 4 A schematic diagram of the automatic measurement and control intelligent box transformer control method in embodiment 1. DETAILED DESCRIPTION

[0052] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0053] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0054] Secondly, the "one embodiment" or "embodiment" referred to herein means that a specific feature, structure or characteristic can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or mutually exclusive of other embodiments.

[0055] Embodiment 1, reference Figs. 1 to 4 For the first embodiment of the present application, the embodiment provides an intelligent box transformer control method for automatic measurement and control, comprising the following steps:

[0056] S1, collecting multi-source data using intelligent sensors, for each intelligent sensor, simulating a group of virtual neurons, calculating activation values, summing the activation values, generating aggregated features, calculating normalized mutual information between the aggregated features, using Chebyshev polynomials to calculate convolution feature vectors, and generating a reduced dimension feature matrix;

[0057] Specifically, collecting multi-source data using intelligent sensors, for each intelligent sensor, simulating a group of virtual neurons, calculating activation values, summing the activation values, generating aggregated features, calculating normalized mutual information between the aggregated features, using Chebyshev polynomials to calculate convolution feature vectors, and generating a reduced dimension feature matrix, including:

[0058] Collecting multi-source data using intelligent sensors, including pressure, current, voltage and temperature data;

[0059] Sorting the multi-source data in chronological order to generate a time series and performing standardization processing;

[0060] The intelligent sensor includes a pressure sensor, a current sensor, a voltage sensor and a temperature sensor;

[0061] Using a low-pass filter to denoise the multi-source data, using the interquartile range method to identify and delete outliers, and using the spline interpolation method to fill in missing data;

[0062] For each intelligent sensor, the sample entropy of the normalized time series is calculated using the standard sample entropy algorithm, and is sorted in ascending order. The quantile rank of the sample entropy in all sample entropies is calculated using the quantile mapping method, and the normalized quantile is obtained. The number of neurons of the intelligent sensor is calculated by linearly mapping the normalized quantile to the range of the number of neurons, and the formula is:

[0063] ,

[0064] wherein is the number of neurons of the i th intelligent sensor, and are the minimum and maximum number of neurons (such as 5 and 10) respectively, is the normalized quantile of the i th intelligent sensor, , is the quantile rank of the sample entropy, and m is the number of intelligent sensors;

[0065] The autocorrelation function value of the intelligent sensor is calculated using the autocorrelation function formula, the first local maximum value (peak value) in the autocorrelation function value is identified as the main period using peak detection, and the time constant is calculated using linear scaling adjustment;

[0066] For each intelligent sensor, the population of virtual neurons is simulated, and the activation value is calculated, and the formula is:

[0067] ,

[0068] wherein is the activation value of the i th intelligent sensor, the j th neuron at time t, is the normalized multi-source data at time h, is the time constant of the i th intelligent sensor, controlling the decay rate;

[0069] The activation value is summed to generate the aggregated feature, and the formula is:

[0070] ,

[0071] wherein is the aggregated feature of the i th intelligent sensor at time t;

[0072] The intelligent sensor is defined as a node, the connection between the nodes is defined as an edge, an undirected graph is constructed, the normalized mutual information between the aggregated features is calculated, and is defined as the edge weight, and the formula is:

[0073] ,

[0074] wherein is the edge weight of the i th and l th nodes, aggregated features between the nodes (calculated using the mutual information formula), and are the aggregated features of the i-th and l-th nodes, respectively;

[0075] The split threshold is set using the percentile method, the edge weights greater than the split threshold are screened, the adjacency matrix of the graph is constructed, the sum of the corresponding row of the node in the adjacency matrix is calculated, and the degree of the node is obtained. The normalized Laplacian matrix is calculated, and the formula is:

[0076] ,

[0077] where is the normalized Laplacian matrix, which describes the topological structure of the graph, D is the degree matrix (constructed based on the degree of the node), W is the adjacency matrix (constructed based on the edge weight), and U is the unit matrix, which ensures the diagonal adjustment of the Laplacian matrix.

[0078] The normalized Laplacian matrix is linearly transformed, and the scaled Laplacian matrix is calculated, and the formula is:

[0079] ,

[0080] where is the scaled Laplacian matrix, is the largest eigenvalue of the normalized Laplacian matrix (estimated by the power iteration method);

[0081] Based on the scaled Laplacian matrix, the Chebyshev polynomial matrix is recursively calculated, and the formula is:

[0082] , ,

[0083] ,

[0084] where is the initial Chebyshev polynomial matrix, is the first-order Chebyshev polynomial matrix, is the k-th order Chebyshev polynomial matrix;

[0085] The convolution feature vector is calculated using the Chebyshev polynomial, and the convolution feature matrix is assembled using the matrix assembly, and the formula is:

[0086] ,

[0087] where is the convolution feature vector of the i-th intelligent sensor at time t, K is the number of polynomial orders, is the k-th order coefficient, which is a scalar;

[0088] The sample covariance matrix method is used to calculate the covariance matrix of the convolution feature matrix, the random singular value decomposition is used to calculate the eigenvalues and eigenvectors of the covariance matrix, the eigenvectors corresponding to the first o largest eigenvalues are selected to form a projection matrix, and the projection matrix is used to project the convolution feature matrix into a low-dimensional space (linear projection is used for projection), to generate a dimensionality reduction feature matrix and perform normalization processing.

[0089] Intelligent sensors are used to collect multi-source data (such as pressure, current, voltage, and temperature), and the data is standardized. Standardization not only makes the data comparable in different dimensions, but also provides consistent input for subsequent analysis. Noise and outliers often affect the accuracy of data and analysis results. De-noising and outlier processing can improve the reliability of data. By mapping sample entropy to the number of neurons, the complexity of sensor data can be converted into a quantitative value. The autocorrelation function is used to analyze the periodicity of time series data, and peak detection helps to determine the main period. Networked data structure can reveal potential associations between different sensors, thereby improving the understanding of the overall behavior of the system. Using normalized Laplacian matrix and Chebyshev polynomial, effective feature extraction and dimensionality reduction can be achieved in high-dimensional data. The system can find key features in high-dimensional data, reduce computational complexity, and maintain important information, thereby improving the accuracy of control strategies. By calculating the covariance matrix and using singular value decomposition for dimensionality reduction, the most representative features can be extracted from complex multi-source data.

[0090] S2, using the square root rule to set the window size, for each window, defining the position vector, using the particle interaction model to calculate the velocity of the window, calculating the anomaly score of the window, using the KL divergence formula to calculate the KL divergence of adjacent windows, using the weighted sum to define the objective function, and generating the control sequence;

[0091] Specifically, the square root rule is used to set the window size, for each window, the position vector is defined, the particle interaction model is used to calculate the velocity of the window, and the anomaly score of the window is calculated, including:

[0092] The square root rule is used to set the window size, the normalized dimensionality reduction feature matrix is segmented according to the window size, the time series slice is used to extract the subsequence, if the last window is less than the window size, it is directly discarded, and the mean and standard deviation of the subsequence are calculated as the statistical characteristics of the window, representing the state intensity of the virtual organism, for each window, the position vector is defined , , and are the mean and standard deviation of the subsequence in the u-th dimension and the v-th window, is the position vector of the u-th dimension in the v-th window. is a two-dimensional real space, representing the mathematical space of position vectors (each vector consists of two real numbers);

[0093] Based on the position vectors, the velocity of the window is calculated using a particle interaction model (combining Euclidean distance (attraction) and exponential decay function (repulsion) to simulate the dynamic interaction between particles), the formula is:

[0094] ,

[0095] ,

[0096] where is the repulsion distance of the vth window, A is the number of dimension reduction dimensions, and are the position vectors of the uth and mth dimensions in the vth window, is the velocity of the uth dimension in the vth window, is a constant;

[0097] The position vector is updated using particle swarm optimization, the formula is:

[0098] ,

[0099] where is the updated position vector, is the step size (set using empirical rules);

[0100] The mean of the updated position vector is calculated, defined as the group center, the anomaly score of the window is calculated using the mean absolute deviation, and normalized, the formula is:

[0101] ,

[0102] where is the anomaly score of the vth window, is the group center.

[0103] The square root rule is used to set the window size and segment the normalized dimensionality reduction feature matrix, which can effectively avoid the calculation complexity caused by too large time window, while ensuring that the small window does not lose the time sequence of the data, through the particle interaction model, the interaction between the system states can be better reflected, so that the control process is more flexible and real-time, the PSO method has strong global search ability and less local optimal problem, which can find the optimal solution in the complex control system, and can greatly improve the stability and efficiency of the system, through the dynamic calculation based on the group center, the effective monitoring of the overall abnormal state of the system can be realized, the greatest advantage is that it can distinguish small system disturbance and large system anomaly, so as to ensure that the control system can maintain accurate monitoring ability under real-time dynamic.

[0104] Further, the KL divergence of adjacent windows is calculated using the KL divergence formula, the target function is defined using the weighted sum, and the control sequence is generated, including:

[0105] The normalized abnormal score is defined as the mean of Gaussian distribution, the standard deviation is calculated using the empirical standard deviation, the KL divergence of adjacent windows is calculated using the KL divergence formula, (in the special case v=1, set , ) formula is:

[0106] ,

[0107] Among them is the KL divergence of the vth and v−1th window, and are the mean and standard deviation of the Gaussian distribution of the vth window, is derived by integrating the probability density function;

[0108] The current state parameters are obtained using the standard industrial protocol, including the switch state (0 represents the closed state, 1 represents the open state), the load distribution ratio and the standby power usage ratio, the load distribution ratio and the standby power usage ratio are normalized, the current state parameters are defined as control parameter vectors using vector construction, and the control parameter vectors are assigned initial values using default values;

[0109] The target function is defined using the weighted sum, and the formula is:

[0110] ,

[0111] Among them is the target function value of the vth window, is the adaptive weight (set using the weighted average method);

[0112] The numerical gradient of the control parameter vector is calculated using the central difference method, the control parameter vector is updated using the gradient descent method based on the numerical gradient, the final control vector is obtained, and each final control vector is connected in time window order to generate the control sequence.

[0113] The use of KL divergence can effectively capture the dynamic changes of the system, especially when detecting system state mutations or trends, KL divergence can provide strong numerical support, by defining the objective function through weighting and optimization through gradient descent method, it can ensure that the control parameters of the system are always in the optimal state. This method can adaptively adjust the control strategy according to real-time data in high dynamic and complex control systems, ensuring that the system always runs on the optimal path, especially in scenarios with large load changes and high environmental uncertainty, it has a significant advantage. The gradient descent method is used to update the control parameter vector, and the control strategy is calculated and optimized through the numerical gradient, which can achieve optimal control effect and avoid excessive control response, thereby improving the stability and robustness of the system. The final control instruction execution ensures the continuity and coordination of the entire control system, and the control strategy can be adjusted in real time to respond to different operating conditions and environmental changes.

[0114] S3, the control sequence generates control instructions and executes, collects feedback data for monitoring and optimization;

[0115] Specifically, the control sequence generates control instructions and executes, including:

[0116] The control sequence is converted into control instructions using a control instruction mapping method;

[0117] The generated control instructions are transmitted and executed through the communication interface Modbus.

[0118] The mapping strategy ensures the efficiency and operability of the control system, Modbus has strong fault tolerance and reliability, and can ensure long-term stable operation in industrial environments. The feedback data is connected in time sequence, which not only improves the continuity of the data, but also monitors the system changes in different time periods in real time, ensuring the stability of the system under various operating conditions.

[0119] Further, the feedback data is collected for monitoring and optimization, including:

[0120] The feedback data is connected in time window order to generate an actual state sequence;

[0121] The error between the control sequence and the actual state sequence is calculated using the absolute error value method;

[0122] The judgment threshold is set using a statistical analysis method, the error is compared with the judgment threshold, the error rate greater than or equal to the judgment threshold is adjusted using a PID control algorithm, and the adjustment is stopped when the difference is less than the judgment threshold, and the feedback data is continuously monitored.

[0123] By calculating the deviation between the control instruction and the actual state, the system can accurately identify whether there is a system deviation and make corresponding adjustments. The advantage of using statistical analysis is that it can not only make reasonable predictions through historical data, but also optimize the control strategy of the system through comprehensive analysis of data distribution, thereby improving the response speed and stability of the system. The system can adjust the control parameters according to the different degrees of deviation to minimize the error and improve the stability of the system. Continuous monitoring provides real-time data for the system, ensuring that the control system maintains stable operation in a changing environment and timely captures potential problems.

[0124] S4, storing the collected and analyzed multi-source data;

[0125] Specifically, the collected and analyzed multi-source data is stored, including:

[0126] The collected multi-source data and the analyzed control sequence are stored in the central database, and security access measures are set. The central database stores the data in the cloud backup, and periodically checks the integrity of the stored data and the backup data. After the detection is completed, an integrity detection record is generated and stored in the central database.

[0127] All data is stored in a central database, avoiding the management difficulty caused by scattered data storage and improving the management efficiency of the system. Through encryption and permission control, only specific personnel can access sensitive data, reducing the risk of data leakage. Cloud backup allows data to be quickly recovered in the event of a disaster, reducing the loss caused by data loss or damage. Through regular integrity detection, it can ensure that the data is not tampered with or lost during storage and transmission, improving the credibility of the data. The storage of detection records enables the system to have an audit function, allowing historical detection results to be traced back, providing support for fault troubleshooting, problem diagnosis and compliance review.

[0128] The embodiment also provides an intelligent box transformer control system for automatic measurement and control, comprising:

[0129] The collection dimension reduction module is used to collect multi-source data using intelligent sensors, simulate a group of virtual neurons for each intelligent sensor, calculate activation values, sum the activation values, generate aggregated features, calculate normalized mutual information between the aggregated features, calculate convolution feature vectors using Chebyshev polynomials, and generate a dimension reduction feature matrix.

[0130] a score control module configured to set window size using square root rule, define position vector for each window, calculate velocity of the window using particle interaction model, calculate abnormal score of the window, calculate KL divergence of adjacent windows using KL divergence formula, define objective function using weighted sum, and generate control sequence;

[0131] an execution monitoring module configured to generate control instructions using the control sequence and perform the control instructions, and collect feedback data for monitoring and optimization;

[0132] a storage module configured to store the collected and analyzed multi-source data.

[0133] The embodiment also provides a computer device suitable for the intelligent box transformer control method of automatic measurement and control, which comprises a memory and a processor.

[0134] The computer device can be a terminal, which comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can also be an external keyboard, touchpad or mouse, etc.

[0135] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the intelligent box transformer control method for realizing automatic measurement and control proposed in the above embodiment; the storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0136] To sum up, the application collects multi-source data by using intelligent sensors, simulates a group of virtual neurons for each intelligent sensor, calculates an activation value, sums the activation values, generates an aggregated feature, calculates normalized mutual information between the aggregated features, calculates a convolution feature vector using a Chebyshev polynomial, and generates a dimension-reduced feature matrix; uses a square root rule to set a window size, defines a position vector for each window, calculates a velocity of the window using a particle interaction model, calculates an anomaly score of the window, calculates a KL divergence of adjacent windows using a KL divergence formula, defines an objective function using a weighted sum, and generates a control sequence; improves the accuracy and efficiency of feature extraction, enhances the correlation analysis between data, and improves the accuracy of the control sequence.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application but not limit the application, although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. An automatic measurement and control intelligent box transformer control method, characterized in that: comprising, collecting multi-source data using smart sensors, for each smart sensor, simulating a population of virtual neurons, calculating activation values, summing the activation values, generating aggregated features, calculating normalized mutual information between the aggregated features, calculating convolution feature vectors using Chebyshev polynomials, generating a reduced dimension feature matrix; setting window size using square root rule, for each window, defining a position vector, calculating the velocity of the window using a particle interaction model, calculating the anomaly score of the window, calculating the KL divergence of adjacent windows using the KL divergence formula, defining the objective function using the weighted sum, generating the control sequence; generating control instructions and executing the control sequence, collecting feedback data for monitoring and optimization; storing the collected and analyzed multi-source data; the collecting multi-source data using smart sensors, for each smart sensor, simulating a population of virtual neurons, calculating activation values, summing the activation values, generating aggregated features, calculating normalized mutual information between the aggregated features, calculating convolution feature vectors using Chebyshev polynomials, generating a reduced dimension feature matrix, comprising: collecting multi-source data using smart sensors, including pressure, current, voltage and temperature data; sorting the multi-source data in chronological order to generate a time series and performing standardization processing; for each smart sensor, calculating the sample entropy of the standardized time series using the standard sample entropy algorithm, and sorting in ascending order, calculating the quantile rank of the sample entropy in all sample entropies using the quantile mapping method, obtaining the normalized quantile, linearly mapping the normalized quantile to the neuron number range, and calculating the neuron number of the smart sensor; calculating the autocorrelation function value of the smart sensor using the autocorrelation function formula, identifying the first local maximum value in the autocorrelation function value as the main period using peak detection, and calculating the time constant using linear scaling adjustment; for each smart sensor, simulating a population of virtual neurons, calculating activation values, summing the activation values, generating aggregated features; defining the smart sensor as a node, defining the connection between the nodes as an edge, constructing an undirected graph, calculating the normalized mutual information between the aggregated features, and defining it as an edge weight; setting the segmentation threshold using the percentile method, filtering the edge weights greater than the segmentation threshold, constructing the adjacency matrix of the graph, calculating the sum of the corresponding rows of the node in the adjacency matrix to obtain the degree of the node, calculating the normalized Laplacian matrix, performing linear transformation on the normalized Laplacian matrix, calculating the scaling Laplacian matrix, and recursively calculating the Chebyshev polynomial matrix based on the scaling Laplacian matrix; calculating convolution feature vectors using Chebyshev polynomials, and assembling convolution feature vectors into a convolution feature matrix using matrix assembly; calculating the covariance matrix of the convolution feature matrix using the sample covariance matrix method, calculating the eigenvalues and eigenvectors of the covariance matrix using random singular value decomposition, selecting the eigenvectors corresponding to the first o largest eigenvalues to form a projection matrix, projecting the convolution feature matrix to a low-dimensional space using the projection matrix, generating a reduced dimension feature matrix and performing normalization processing.

2. The method of claim 1, wherein: The square root rule is used to set the window size, for each window, the position vector is defined, the particle interaction model is used to calculate the velocity of the window, the anomaly score of the window is calculated, including: Using the square root rule to set the window size, the normalized dimensionality reduction feature matrix is segmented according to the window size, the time series slice is used to extract the subsequence, and the mean and standard deviation of the subsequence are calculated as the statistical characteristics of the window. For each window, define the position vector ; Based on the position vector, the particle interaction model is used to calculate the velocity of the window, the position vector is updated using particle swarm optimization; Calculate the mean of the updated position vector, defined as the group center, calculate the anomaly score of the window using the mean absolute deviation, and perform normalization processing.

3. The method of claim 2, wherein: The KL divergence formula is used to calculate the KL divergence of adjacent windows, and the weighted sum is used to define the objective function to generate the control sequence, including: The normalized anomaly score is defined as the mean of the Gaussian distribution, the standard deviation is calculated using the empirical standard deviation, and the KL divergence formula is used to calculate the KL divergence of adjacent windows; The current state parameters are obtained using the standard industrial protocol, the current state parameters are defined as the control parameter vector using the vector structure, and the control parameter vector is assigned the initial value using the default value; The weighted sum is used to define the objective function, the numerical gradient of the control parameter vector is calculated using the central difference method, the control parameter vector is updated based on the numerical gradient using the gradient descent method, and the final control vector is obtained. Connect each final control vector in time window order to generate a control sequence.

4. The method of claim 3, wherein: The control sequence is generated into a control instruction and executed, including: The control sequence is converted into a control instruction using the control instruction mapping method; The generated control instruction is transmitted and executed through the communication interface Modbus.

5. The method of claim 4, wherein: The collected feedback data is monitored and optimized, including: The feedback data is connected in time window order to generate an actual state sequence; The error between the control sequence and the actual state sequence is calculated using the absolute error value method; Use statistical analysis method to set judgment threshold, compare error with judgment threshold, use PID control algorithm to adjust error rate greater than or equal to judgment threshold, stop adjusting until difference is less than judgment threshold, continue to monitor feedback data.

6. The method of claim 5, wherein: The collected and analyzed multi-source data is stored, including: The collected multi-source data and the control sequence generated by analysis are stored in the central database, and security access measures are set. The central database stores the data in the cloud backup, and regularly checks the integrity of the stored data and the backup data. After the detection is completed, the integrity detection record is stored in the central database.

7. An automatic measurement and control intelligent box transformer control system based on the automatic measurement and control intelligent box transformer control method of any one of claims 1-6, characterized in that: Including, The collection dimension reduction module is used to collect multi-source data using intelligent sensors, simulate a group of virtual neurons for each intelligent sensor, calculate the activation value, sum the activation value, generate aggregated features, calculate the normalized mutual information between aggregated features, calculate the convolution feature vector using Chebyshev polynomial, and generate the dimension reduction feature matrix; The score control module is used to set the window size using the square root rule, define the position vector for each window, calculate the velocity of the window using the particle interaction model, calculate the anomaly score of the window, calculate the KL divergence of adjacent windows using the KL divergence formula, and generate the control sequence using the weighted sum to define the objective function. The execution monitoring module is used to generate control instructions from the control sequence and execute them, and monitor and optimize the collected feedback data. A storage module is configured to store the collected and analyzed multi-source data.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the intelligent box transformer control method of automatic measurement and control according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the intelligent box transformer control method of automatic measurement and control according to any one of claims 1-6.

Images