Intelligent monitoring and regulating system for connector

Abstract

The invention discloses an intelligent monitoring and regulation system for a connector, and relates to the field of monitoring data processing, and the key points of the technical scheme are that a data acquisition module acquires comprehensive working data and environmental data of the connector; the data optimization module is used for preprocessing the comprehensive working data, and the preprocessing modes comprise data synchronous calibration, weighting processing on repeated current values or voltage values existing at a single time point, and interpolation method filling on data missing at the single time point; obtaining adjustment data; the feature analysis module is used for performing singular spectrum analysis SSA on the adjustment and calibration data to obtain feature data; and the prediction regulation and control module inputs the feature data into a pre-trained neural network model, a regulation and control decision of the connector is obtained through prediction, all the modules are connected in a wired and / or wireless mode, and intelligent monitoring and regulation and control of the connector are achieved.

Description

Technical Field

[0001] The present invention relates to the field of monitoring data processing, and more specifically, to a connector intelligent monitoring and control system. Background Art

[0002] A connector is a device used to connect two or more electronic devices or components. It is an interface for transmitting power signals or data. Connectors usually consist of a plug (socket) and a socket (slot), which connect the power data signals between devices through a specific physical interface. Connectors can have various shapes, sizes and types depending on their purpose and design.

[0003] With the rapid development of modern cities and industries, the amount of data faced by power safety supervision has shown explosive growth. Connector supervision data is not only huge in quantity, but the rapid growth of data volume also brings processing pressure, which increases the difficulty of extracting valuable information from the data. It is difficult to quickly respond to and extract the real-time status of the connector. Through the monitoring of various intelligent sensors, the sources of connector supervision data are diversified, which makes the structure diverse, and the data integration work becomes particularly complicated, which makes data analysis work extremely complicated, resulting in a lack of support for connector regulation and prolonged decision-making time, making it difficult to provide real-time regulation feedback for abnormal situations. Therefore, it is necessary to design intelligent monitoring data analysis for connectors. Summary of the invention

[0004] In view of the problems existing in the prior art, the purpose of the present invention is to provide a connector intelligent monitoring and control system to achieve safe monitoring of campus data.

[0005] To achieve the above object, the present invention provides the following technical solution: a connector intelligent monitoring and control system, the connector intelligent monitoring and control system comprising:

[0006] Data collection module: collects the comprehensive working data and environmental data of the connector;

[0007] Data optimization module: used to pre-process comprehensive working data and obtain adjustment data;

[0008] Feature analysis module: used to perform singular spectrum analysis SSA on the calibration data to obtain feature data;

[0009] Prediction and control module: inputs feature data into a pre-trained neural network model to predict the control decision of the connector;

[0010] The modules are connected to each other via wired and / or wireless means.

[0011] Preferably, the comprehensive working data includes current data and voltage data, the current data includes a set of current values ​​within a cycle time, the voltage data includes a set of voltage values ​​within a cycle time, and the environmental data includes temperature data and humidity data.

[0012] Preferably, the preprocessing method includes data synchronization calibration, and the data synchronization calibration specifically includes:

[0013] Construct a unified time axis: define the acquisition time period, identify the timestamp, use the start and end time of the current data and voltage data acquisition within the time period as the time range of the unified time axis, and use the time series generated with a fixed time interval t as the time point;

[0014] Align current data and voltage data based on a unified time axis: traverse the timestamps of each current value and voltage value in the current data and voltage data, calculate the difference between the timestamp of each current value and voltage value and all time points on the unified time axis, and select the time point corresponding to the minimum difference for insertion. Perform weighted processing on repeated current values ​​or voltage values ​​at a single time point, and fill in missing data at a single time point using interpolation.

[0015] Generate calibration data: correspond each time point on the unified time axis to a current value and a voltage value respectively, and integrate all current data and voltage data into current synchronization data and voltage synchronization data according to the aligned time points respectively. The current synchronization data and voltage synchronization data together constitute the calibration data.

[0016] Preferably, the weighted processing of repeated current values ​​or voltage values ​​at a single time point includes:

[0017] The current value or voltage value is set to vi, i represents the index of the current value or voltage value at a single time point, and i=1, ..., I, I is the number of current values ​​or voltage values ​​at a single time point, and the current value or voltage value inserted at a single time point is updated;

[0018]

[0019] Among them, v New represents the current value or voltage value inserted into a single time point after weighted processing; ωi represents the weight;

[0020] The weight ωi is set as:

[0021]

[0022] Among them, si represents the timestamp on vi with index i, and st represents the time point to be inserted.

[0023] Preferably, the interpolation method for filling missing data at a single time point includes:

[0024] Identify the characteristic properties of current data and voltage data, perform improved linear interpolation on current data and voltage data with non-periodic properties, and perform trigonometric function interpolation on current data and voltage data with periodic properties;

[0025] Improved linear interpolation for current data and voltage data with non-periodic properties includes:

[0026] Set the missing data value at the time point s(x) on the unified time axis to v nan x, the data value is used to represent the current value or voltage value, the proportion coefficient λ is introduced, and the missing data value v is calculated nan i,

[0027]

[0028] Among them, v(x-1) represents the data value at the time point s(x-1) before the time point s(x), and v(x+1) represents the data value at the time point s(x+1) after the time point s(x).

[0029] Preferably, the performing trigonometric function interpolation on the current data and voltage data having periodic properties comprises:

[0030] Construct a trigonometric function model: select the trigonometric function y(u)=Asin(Bu+C)+D, determine the values ​​of A, B, C and D through least squares fitting, and obtain a complete trigonometric function expression;

[0031] Supplementation of missing data values: Set the time point where there is a missing data value on the unified time axis as u, calculate the value at time point u according to the constructed trigonometric function model, fill the calculated value into the corresponding time point u, traverse all time points with missing data values, and fill them one by one.

[0032] Preferably, performing singular spectrum analysis SSA on the calibration data comprises:

[0033] Construct analysis matrix: organize the calibration data into a time series data matrix, where rows represent time points and columns represent current values ​​and voltage values, to form an analysis matrix;

[0034] Embed the delay embedding matrix: Construct the delay embedding matrix; define the delay window size L and the data vector length N, set the number of rows of the delay embedding matrix to N-L+1, and use L consecutive data points in the analysis matrix as a subsequence for each column of the delay embedding matrix. Slide along the time axis in sequence to form the rows of the delay embedding matrix, and embed the analysis matrix into the delay embedding matrix. The delay embedding matrix is ​​as follows:

[0035]

[0036] Perform singular value decomposition on the delayed embedding matrix, select the left singular vectors corresponding to the first k singular values ​​with larger singular values, and obtain the principal component matrix. Use the principal component matrix to reconstruct the original time series data matrix to obtain the feature data after the calibration data features are reconstructed.

[0037] Preferably, the training of the neural network model includes:

[0038] Define recursive neural network as the basic structure of neural network model, which includes input layer, hidden layer and output layer;

[0039] Take the time points on the unified time axis as time labels, add the time labels to the feature data, obtain the time series data set, use the time series data set as the training input of the neural network model, and use the control decision of the connector as the prediction output of the neural network model;

[0040] Initialize the network parameters of the neural network model; the network parameters include the weight matrix from the input layer to the hidden layer, the cyclic weight matrix of the hidden layer, the bias vector of the hidden layer, the weight matrix from the hidden layer to the output layer, and the bias vector of the output layer;

[0041] Define the loss function of the neural network model,

[0042]

[0043] Wherein, M is the number of data values ​​in the time series data set, and the data values ​​include current values ​​and voltage values; b is the index of the data value, α and β are weight parameters, and Jb is the actual electrical anomaly judgment value. is the electrical anomaly judgment value output by the model prediction, Eh is the actual environmental impact coefficient, Environmental impact coefficients for model prediction output;

[0044] Get the sample set, set the time series data set as a subset, divide the sample set into multiple batches, pass each batch of the time series data set to the input layer in turn, update the network parameters according to the optimization algorithm, and stop when the loss function value is minimized and no longer changes, thus completing the training of the neural network model.

[0045] Preferably, the electrical abnormality determination value;

[0046]

[0047] Environmental impact factor,

[0048]

[0049] Among them, r1, r2, r3, r4, r5 and r6 are weight parameters, which can be set by the management personnel according to experimental analysis or experience; Tt is the real-time temperature value obtained during the cycle time; T1 is the maximum safe temperature; T2 is the minimum safe temperature; Sd is the real-time humidity value obtained during the cycle time; Dy is the real-time voltage value obtained during the cycle time; Dl is the real-time current value obtained during the cycle time; is the calibration voltage value; is the calibration current value.

[0050] Preferably, the proportion coefficient λ is calculated as,

[0051]

[0052] Wherein, v(b) represents the standard data value.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] In the present invention, the present invention performs synchronous calibration on the current data and voltage data from different sensor sources on the connector through preprocessing of communication data, ensures the consistency and comparability of the data on a unified time axis, reduces the interference of time misalignment on the electrical data analysis process, reduces the operating computing power of data analysis, and realizes high-quality input of communication data at the same time, improves the effect of subsequent data processing and model training as well as the retrieval and response speed, and provides an accurate data basis for subsequent control strategy analysis and model optimization; realizes feature interpolation of periodic data and non-periodic data through interpolation method, fully retains the characteristics of data in the process of data supplementation, provides a good calculation basis for singular spectrum analysis SSA, accelerates the efficiency of feature extraction, and injects new kinetic energy into the agility and flexibility of data processing. The intelligent control of connectors based on neural network models realized by the system intelligently outputs control decisions for different states (abnormal and non-abnormal) of connectors; the adaptive control optimization mechanism continuously improves the overall efficiency of data processing, and multiple modules cooperate with each other to realize comprehensive data collection, processing and analysis, provide real-time and accurate data support, and help managers better understand the operation of connectors and make corresponding control countermeasures. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of the structure of a connector intelligent monitoring and control system is proposed for the present invention;

[0056] Figure 2 Schematic diagram of the method steps applied to the connector intelligent monitoring and control system in the present invention. DETAILED DESCRIPTION

[0057] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0058] Embodiment 1

[0059] Reference Figure 1 and Figure 2 , Example 1 further illustrates a connector intelligent monitoring and control system proposed by the present invention.

[0060] A connector is a device used to connect two or more electronic devices or components. It is an interface for transmitting power signals or data. Connectors usually consist of a plug (socket) and a socket (slot), which connect the power data signals between devices through a specific physical interface. Connectors can have various shapes, sizes and types depending on their purpose and design.

[0061] With the rapid development of modern cities and industries, the amount of data faced by power safety supervision has shown explosive growth. Connector supervision data is not only huge in quantity, but the rapid growth of data volume also brings processing pressure, which increases the difficulty of extracting valuable information from the data. It is difficult to respond and extract the real-time status of the connector quickly. Through the monitoring of various intelligent sensors, the sources of connector supervision data are diversified, which makes the structure diverse, and data integration work becomes particularly complicated, which makes data analysis work extremely complicated, resulting in a lack of support for connector regulation and prolonged decision-making time, making it difficult to provide real-time regulation feedback for abnormal situations.

[0062] The connector intelligent monitoring and control system is a system used to monitor and manage the performance and status of connectors. This system can monitor the connector's connection status, electrical performance, temperature, humidity and other parameters in real time, and automatically control and alarm according to preset conditions.

[0063] Data collection module: collects the comprehensive working data and environmental data of the connector.

[0064] Install an appropriate number and type of sensors (such as voltage sensors, current sensors, temperature sensors, humidity sensors, etc.) to collect various parameter data of the connector.

[0065] The comprehensive working data includes current data and voltage data. The current data includes a collection of current values ​​within a cycle time. The voltage data includes a collection of voltage values ​​within a cycle time. The environmental data includes temperature data and humidity data. The current data and voltage data are obtained through current sensors and voltage sensors. The temperature data and humidity data are obtained through temperature sensors and humidity sensors.

[0066] The current data can reflect the current current load of the connector. If the current exceeds the rated capacity of the connector, it may cause the connector to be overloaded, causing failure or damage. The voltage data can show the voltage stability and fluctuation of the connector. Unstable voltage may cause poor performance or malfunction of the connector. The temperature data can indicate whether the operating temperature range of the connector is normal. Too high or too low temperature may affect the performance and life of the connector. For example, high temperature may cause the expansion and failure of the connector material, and low temperature may cause the connector material to become brittle. The humidity data can reflect the humidity level of the environment in which the connector is located. High humidity may cause corrosion in the connector or moisture in the insulating material, thereby reducing its performance and life.

[0067] Data optimization module: used to pre-process comprehensive working data and obtain adjustment data.

[0068] Data synchronization calibration is a very important step in the connector intelligent monitoring and control system, which is used to ensure the accuracy and consistency of the collected data. Data calibration refers to the correction of the collected raw data to eliminate the deviation caused by sensor error. Calibration can be performed according to known standards or reference data to ensure the accuracy and reliability of the data.

[0069] Data synchronization calibration refers to the time alignment of data from different sensors to ensure that they have the same timestamp or sampling interval. In the connector system, different sensors may have different sampling rates or operating frequencies, and data synchronization processing is required to ensure data consistency and comparability.

[0070] The data collected by the sensor contains a timestamp, which can be calibrated when the sensor is installed and configured to ensure that the data contains timestamp information. During the calibration process, the correspondence between the timestamp and the sensor reading can be recorded.

[0071] In the present invention, for the comprehensive working data collected by the sensors, if they use different timestamp formats, scripts can be written or data processing tools can be used to convert them into a unified format. The scripts or data processing tools can be the datetime module in Python or the Date object in JavaScript, which can easily convert and process the timestamp format and convert it into a target format. A standard timestamp format is selected as the target format, such as the ISO8601 format (such as YYYY-MM-DDTHH:MM:SS).

[0072] Data synchronization calibration specifically includes:

[0073] Build a unified time axis: define the acquisition time period, identify the timestamp, and use the start and end time of the current data and voltage data acquisition within the time period as the time range of the unified time axis. The start and end time can be based on the time range of the current data or voltage data, or the earliest start time and the latest end time of all data can be used as the time point to generate a time series with a fixed time interval t (t can be assigned a value, such as 2 seconds or every millisecond);

[0074] Align current data and voltage data based on a unified time axis: traverse the timestamps of each current value and voltage value in the current data and voltage data, calculate the difference between the timestamp of each current value and voltage value and all time points on the unified time axis, and select the time point corresponding to the minimum difference for insertion. Perform weighted processing on repeated current values ​​or voltage values ​​at a single time point, and fill in missing data at a single time point using interpolation.

[0075] For each current value and voltage value corresponding to the timestamp, find its closest time point on the unified time axis. The selection of the closest time point can be achieved by calculating the difference between the timestamp and all time points on the unified time axis and selecting the time point corresponding to the minimum difference.

[0076] Each current value and voltage value is aligned to the nearest time point on the unified time axis, and the data is inserted into the time point generated by each time series interval on the unified time axis to ensure that each current value and voltage value in the current data and voltage data is aligned in time.

[0077] Generate calibration data: Each time point on the unified time axis corresponds to a current value and a voltage value. All current data and voltage data are integrated according to the aligned time points to form current synchronization data and voltage synchronization data. The current synchronization data and voltage synchronization data together constitute the calibration data, ensuring the temporal consistency and alignment of data collected by different sensors.

[0078] The weighted processing for repeated current or voltage values ​​at a single time point includes:

[0079] The current value or voltage value is set to vi, i represents the index of the current value or voltage value at a single time point, and i=1, ..., I, I is the number of current values ​​or voltage values ​​at a single time point, and the current value or voltage value inserted at a single time point is updated;

[0080]

[0081] Among them, v New represents the current value or voltage value inserted into a single time point after weighted processing; ωi represents the weight;

[0082] The weight ωi is set as:

[0083]

[0084] Where si represents the timestamp on vi with index i, and st represents the time point to be inserted;

[0085] When the timestamps of multiple current values ​​and voltage values ​​in the current data and voltage data are closest to the same time point on the unified time axis, the repeated timestamps can be processed by weighted processing. The following is a simple embodiment, in which it is assumed that the timestamps of two current values ​​coincide with a certain time point on the unified time axis, and the weighted processing is as follows:

[0086] From this we can see that i = 1, 2, the data at a certain time point is represented as v1, v2, and the weighted data is obtained

[0087] In practical applications, weights can be set based on the reliability, importance or other factors of the data source to achieve a reasonable weighting effect. Through weighted processing, we can merge multiple data points that are repeated in time into one data point, thereby ensuring consistency on a unified timeline.

[0088] Interpolation methods for missing data at a single time point include:

[0089] If the timestamps of some data sources are not completely aligned with the time points on the unified timeline, consider using interpolation methods to fill in the missing data points. The periodic characteristics of the current and voltage measured in the connector depend on the characteristics of the circuit or device to which the connector is connected. In general, the current and voltage measured in the connector may be periodic, non-periodic, or mixed:

[0090] If the circuit connected to the connector is an AC circuit, such as an AC power supply, AC motor, or AC lamp, the measured current and voltage usually have periodic characteristics. In this case, the current and voltage change periodically over time, and their waveforms can be sinusoidal or approximately sinusoidal.

[0091] If the circuit to which the connector is connected is a DC circuit or a non-periodic circuit, such as a battery-powered device, a DC power supply, or some non-linear circuit, the measured current and voltage may not have periodic characteristics. In this case, the current and voltage are usually constant or change in a non-periodic manner.

[0092] In some cases, the current and voltage measured in the connector may be of mixed nature. For example, the connector may be connected to both AC and DC power supplies, or to devices with periodic and non-periodic signals. In this case, the measured current and voltage may have different periodic characteristics or be non-periodic.

[0093] Therefore, whether the current and voltage measured in the connector are periodic depends on the nature of the connected circuit or device, and needs to be analyzed and judged according to the specific situation. When interpolating missing data, first determine the load connected when the sensor monitors the connector power data, and then select the interpolation method by identifying the AC and DC characteristics of the load.

[0094] Identify the characteristic properties of current data and voltage data, perform improved linear interpolation on current data and voltage data with non-periodic properties, and perform trigonometric function interpolation on current data and voltage data with periodic properties;

[0095] Improved linear interpolation for current data and voltage data with non-periodic properties includes:

[0096] Set the missing data value at the time point s(x) on the unified time axis to v nan x, the data value is used to represent the current value or voltage value, the proportion coefficient λ is introduced, and the missing data value v is calculated nan i,

[0097]

[0098] Among them, v(x-1) represents the data value at the time point s(x-1) before the time point s(x), and v(x+1) represents the data value at the time point s(x+1) after the time point s(x).

[0099] Calculate the proportion coefficient λ,

[0100]

[0101] Among them, v(b) represents the standard data value, which can be set by the management personnel based on experience. When there are multiple consecutive missing data values, the standard data values ​​can be inserted at intervals, and then the remaining missing data values ​​can be calculated.

[0102] The trigonometric interpolation of current data and voltage data with periodic properties includes:

[0103] Construct a trigonometric function model: select the trigonometric function y(u)=Asin(Bu+C)+D, determine the values ​​of A, B, C and D through least squares fitting, and obtain a complete trigonometric function expression;

[0104] Supplementation of missing data values: Set the time point where there is a missing data value on the unified time axis as u, calculate the value at time point u according to the constructed trigonometric function model, fill the calculated value into the corresponding time point u, traverse all time points with missing data values, and fill them one by one.

[0105] Trigonometric interpolation is a method for fitting periodic data and interpolating based on trigonometric functions (such as sine or cosine functions). According to the periodic characteristics of the data, an appropriate trigonometric function is selected as the interpolation function. The present invention selects the sine function as the interpolation function because the sine function is periodic and can fit periodic data well.

[0106] Use the selected trigonometric function y(u)=Asin(Bu+C)+D to fit the known data values, where A, B, C and D are fitting parameters. The best fitting parameters are found through least squares fitting, and a trigonometric function model is constructed. The fitted trigonometric function model is used to calculate the data value that needs to be inserted at the missing time point. For the insertion time point u, the interpolation result is obtained by calculating the function value of the selected trigonometric function model at u, and the calculated interpolation value is filled into the missing position to complete the interpolation processing.

[0107] The values ​​of A, B, C, and D determined by least squares fitting include:

[0108] Define the fitting function: Use the sine function as the fitting function, A, B, C and D are the fitting parameters, and its general form is f(g; A, B, C, D) = Asin(Bg + C) + D, where g represents the variable, i.e. the data value brought in;

[0109] Definition of error function: The error function is used to measure the error between the fitting function and the known data value. The present invention uses the square error function.

[0110]

[0111] Among them, f(g e ,A,B,C,D) represents the fitting function value generated by the e-th data value, y(g e ) represents the e-th known data value, and n represents the number of data values;

[0112] Minimize the error function: The goal of least squares fitting is to minimize the error function. The gradient descent method is used to iteratively adjust the values ​​of parameters A, B, C, and D until the square error function value E(A, B, C, D) reaches a minimum, and A, B, C, and D corresponding to the minimum value are selected as the optimal fitting parameters.

[0113] Feature analysis module: used to perform singular spectrum analysis SSA on the calibration data to obtain feature data;

[0114] Singular spectrum analysis SSA of the calibration data includes:

[0115] Construct analysis matrix: organize the calibration data into a time series data matrix, where rows represent time points and columns represent current values ​​and voltage values, to form an analysis matrix;

[0116] Embed the delayed embedding matrix:

[0117] Construct a delay embedding matrix: define the delay window size L and the data vector length N. The data vector length N is the same as the time points on the unified time axis or the number of current values ​​and voltage values ​​in the calibration data. Set the number of rows of the delay embedding matrix to N-L+1. Each column of the delay embedding matrix takes L consecutive data points in the analysis matrix as a subsequence, and slides along the time axis in sequence to form a row of the delay embedding matrix. Embed the analysis matrix into the delay embedding matrix.

[0118] For example, in the analysis matrix, rows represent time points, columns represent current values ​​and voltage values, and then a time series data of length N is obtained, recorded as h1,...,hN, where hr represents the rth time series data point, and r=1,...,N. Through a sliding window with a delay window size of L, the delay embedding matrix can be constructed as follows:

[0119]

[0120] Singular value decomposition is an important matrix decomposition technique and is widely used in singular spectrum analysis (SSA). It decomposes a matrix into the product of three matrices, thereby extracting the structural information and main features of the matrix. Singular value decomposition (SVD) is a classic matrix decomposition technique that is widely studied and applied in mathematics and engineering. For a given embedded matrix, the existing technology does provide mature SVD algorithms and libraries that can be used directly. Many numerical computing software and programming languages ​​(such as NumPy, SciPy in Python, MATLAB, etc.) provide efficient SVD implementations. These libraries can handle matrices of different sizes and types and return the results of SVD decomposition. By calling the functions in these libraries, the embedded matrix is ​​directly decomposed into a left singular vector matrix, a singular value matrix, and a right singular vector matrix.

[0121] The delayed embedding matrix is ​​subjected to singular value decomposition to obtain a left singular vector matrix, a singular value matrix and a right singular vector matrix. For example, given an m×l real or complex matrix, the left singular vector matrix obtained by SVD decomposition is an m×m unitary matrix, whose column vectors are called left singular vectors; the singular value matrix is ​​an m×l diagonal matrix, and the elements on the diagonal are called singular values, which are usually arranged in descending order; the right singular vector matrix is ​​an l×l unitary matrix, and its row vectors are called right singular vectors.

[0122] Perform singular value decomposition on the delayed embedding matrix,

[0123] X=UΣV T ;

[0124] U represents the left singular vector matrix, Σ represents the singular value matrix, V T represents the right singular vector matrix;

[0125] Select the left singular vectors corresponding to the first k singular values ​​with larger singular values, and get the principal component matrix FF = [F1...Fk]. Use the principal component matrix to reconstruct the original time series data matrix, and get the feature data Y = FF·V after the calibration data feature is reconstructed. T ;

[0126] Among them, the delay window size L and k value can be set according to the management experience, or can be obtained according to experimental analysis;

[0127] Prediction and control module: inputs feature data into a pre-trained neural network model to predict the control decision of the connector;

[0128] The training of the neural network model includes:

[0129] Define recursive neural network as the basic structure of neural network model, which includes input layer, hidden layer and output layer;

[0130] Take the time points on the unified time axis as time labels, add the time labels to the feature data, obtain the time series data set, use the time series data set as the training input of the neural network model, and use the control decision of the connector as the prediction output of the neural network model;

[0131] Initialize the network parameters of the neural network model; the network parameters include the weight matrix from the input layer to the hidden layer, the cyclic weight matrix of the hidden layer, the bias vector of the hidden layer, the weight matrix from the hidden layer to the output layer, and the bias vector of the output layer;

[0132] Define the loss function of the neural network model,

[0133]

[0134] Wherein, M is the number of data values ​​in the time series data set, and the data values ​​include current values ​​and voltage values; b is the index of the data value, α and β are weight parameters, and Jb is the actual electrical anomaly judgment value. is the electrical anomaly judgment value output by the model prediction, Eh is the actual environmental impact coefficient, Environmental impact coefficients for model prediction output;

[0135] Electrical abnormality judgment value;

[0136]

[0137] Environmental impact factor,

[0138]

[0139] Among them, r1, r2, r3, r4, r5 and r6 are weight parameters, which can be set by the management personnel according to experimental analysis or experience; Tt is the real-time temperature value obtained during the cycle time; T1 is the maximum safe temperature; T2 is the minimum safe temperature; Sd is the real-time humidity value obtained during the cycle time; Dy is the real-time voltage value obtained during the cycle time; Dl is the real-time current value obtained during the cycle time; is the calibration voltage value; is the calibration current value;

[0140] Get a sample set, which includes sufficient subsets. Set the time series data set as a subset, divide the sample set into multiple batches, pass each batch of the time series data set to the input layer in turn, and then pass through the hidden layer and the output layer in turn to obtain the predicted output of the model; calculate the value of the corresponding loss function, and backpropagate the error gradient from the output layer to the input layer; according to the error gradient, use the optimization algorithm to update the network parameters to reduce the value of the loss function; stop when the value of the loss function is minimized and no longer changes, that is, the training of the neural network model is completed.

[0141] The time series data set within the current cycle time is input into the trained neural network model to obtain the connector's control decision, and the connector is adjusted and controlled according to the output control decision. The control decision may include adjusting connector parameters, issuing alarms, performing automatic repair actions, etc., to maintain the normal operation of the connector and realize comprehensive monitoring and intelligent control of the connector status.

[0142] In this embodiment, the present invention performs synchronous calibration on the current data and voltage data from different sensor sources on the connector through preprocessing of communication data, ensures the consistency and comparability of the data on a unified time axis, reduces the interference of time misalignment on the electrical data analysis process, reduces the operating computing power of data analysis, and realizes high-quality input of communication data, improves the effect of subsequent data processing and model training, as well as the retrieval and response speed, and provides an accurate data basis for subsequent control strategy analysis and model optimization; the interpolation method is used to realize feature interpolation of periodic data and non-periodic data, fully retains the characteristics of the data in the process of data supplementation, provides a good calculation basis for singular spectrum analysis SSA, accelerates the efficiency of feature extraction, and injects new kinetic energy into the agility and flexibility of data processing. The intelligent control of the connector based on the neural network model realized by the system can intelligently output control decisions for different states (abnormal and non-abnormal) of the connector; the adaptive control optimization mechanism continuously improves the overall efficiency of data processing, and multiple modules cooperate with each other to realize comprehensive data collection, processing and analysis, provide real-time and accurate data support, and help managers better understand the operation of the connector and make corresponding control countermeasures.

[0143] Embodiment 2

[0144] Reference Figure 1 and Figure 2 , Example 2 further illustrates a connector intelligent monitoring and control system proposed by the present invention.

[0145] The connector intelligent monitoring and control method is applied to a connector intelligent monitoring and control system, and comprises the following steps:

[0146] S1. Collect comprehensive working data and environmental data of the connector within the current cycle time;

[0147] S2. Preprocessing the comprehensive working data by using a data synchronization calibration method to obtain preprocessed calibration data;

[0148] S3, analyzing the current data and voltage data in the calibration data by singular spectrum analysis SSA to obtain characteristic data;

[0149] S4, constructing a neural network model and training it to obtain a trained neural network model;

[0150] S5. Import the characteristic data of the current cycle time into the trained neural network model, obtain the control decision of the connector, and perform relevant control.

[0151] The connector intelligent monitoring and control system comprises:

[0152] Data collection module: collects the comprehensive working data and environmental data of the connector;

[0153] Data optimization module: used to pre-process comprehensive working data and obtain adjustment data;

[0154] Feature analysis module: used to perform singular spectrum analysis SSA on the calibration data to obtain feature data;

[0155] Prediction and control module: inputs feature data into a pre-trained neural network model to predict the control decision of the connector;

[0156] The modules are connected to each other via wired and / or wireless means.

[0157] In addition, according to the implementation of the present application, the process described in the figure of a connector intelligent monitoring and control system can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application. Of course, the architecture shown in the figure of a connector intelligent monitoring and control system is only exemplary. When implementing different devices, adaptive selection or adjustment can be performed according to actual needs.

[0158] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0159] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A connector intelligent monitoring and control system, characterized in that: The connector intelligent monitoring and control system includes: Data collection module: collects the comprehensive working data and environmental data of the connector; Data optimization module: used to pre-process comprehensive working data and obtain adjustment data; Feature analysis module: used to perform singular spectrum analysis SSA on the calibration data to obtain feature data; Prediction and control module: inputs feature data into a pre-trained neural network model to predict the control decision of the connector; The modules are connected to each other via wired and / or wireless means.

2. A connector intelligent monitoring and control system according to claim 1, characterized in that: The comprehensive working data includes current data and voltage data. The current data includes a set of current values ​​within a cycle time. The voltage data includes a set of voltage values ​​within a cycle time. The environmental data includes temperature data and humidity data.

3. A connector intelligent monitoring and control system according to claim 2, characterized in that: The preprocessing method includes data synchronization calibration, which specifically includes: Construct a unified time axis: define the acquisition time period, identify the timestamp, use the start and end time of the current data and voltage data acquisition within the time period as the time range of the unified time axis, and use the time series generated with a fixed time interval t as the time point; Align current data and voltage data based on a unified time axis: traverse the timestamps of each current value and voltage value in the current data and voltage data, calculate the difference between the timestamp of each current value and voltage value and all time points on the unified time axis, and select the time point corresponding to the minimum difference for insertion. Perform weighted processing on repeated current values ​​or voltage values ​​at a single time point, and fill in missing data at a single time point using interpolation. Generate calibration data: correspond each time point on the unified time axis to a current value and a voltage value respectively, and integrate all current data and voltage data into current synchronization data and voltage synchronization data according to the aligned time points respectively. The current synchronization data and voltage synchronization data together constitute the calibration data.

4. A connector intelligent monitoring and control system according to claim 3, characterized in that: The weighted processing of repeated current values ​​or voltage values ​​at a single time point includes: The current value or voltage value is set to vi, i represents the index of the current value or voltage value at a single time point, and i=1, ..., I, I is the number of current values ​​or voltage values ​​at a single time point, and the current value or voltage value inserted at a single time point is updated; Among them, v New represents the current value or voltage value inserted into a single time point after weighted processing; ωi represents the weight; The weight ωi is set as: Among them, si represents the timestamp on vi with index i, and st represents the time point to be inserted.

5. A connector intelligent monitoring and control system according to claim 4, characterized in that: The interpolation method for filling missing data at a single time point includes: Identify the characteristic properties of current data and voltage data, perform improved linear interpolation on current data and voltage data with non-periodic properties, and perform trigonometric function interpolation on current data and voltage data with periodic properties; Improved linear interpolation for current data and voltage data with non-periodic properties includes: Set the missing data value at the time point s(x) on the unified time axis to v nan x, the data value is used to represent the current value or voltage value, the proportion coefficient λ is introduced, and the missing data value v is calculated nan i, Among them, v(x-1) represents the data value at the time point s(x-1) before the time point s(x), and v(x+1) represents the data value at the time point s(x+1) after the time point s(x).

6. A connector intelligent monitoring and control system according to claim 5, characterized in that: The performing trigonometric function interpolation on the current data and voltage data having periodic properties comprises: Construct a trigonometric function model: select the trigonometric function y(u)=Asin(Bu+C)+D, determine the values ​​of A, B, C and D through least squares fitting, and obtain a complete trigonometric function expression; Supplementation of missing data values: Set the time point where there is a missing data value on the unified time axis as u, calculate the value at time point u according to the constructed trigonometric function model, fill the calculated value into the corresponding time point u, traverse all time points with missing data values, and fill them one by one.

7. A connector intelligent monitoring and control system according to claim 6, characterized in that: The singular spectrum analysis SSA of the calibration data comprises: Construct analysis matrix: organize the calibration data into a time series data matrix, where rows represent time points and columns represent current values ​​and voltage values, to form an analysis matrix; Embed the delay embedding matrix: Construct the delay embedding matrix; define the delay window size L and the data vector length N, set the number of rows of the delay embedding matrix to N-L+1, and use L consecutive data points in the analysis matrix as a subsequence for each column of the delay embedding matrix. Slide along the time axis in sequence to form the rows of the delay embedding matrix, and embed the analysis matrix into the delay embedding matrix. The delay embedding matrix is ​​as follows: Perform singular value decomposition on the delayed embedding matrix, select the left singular vectors corresponding to the first k singular values ​​with larger singular values, and obtain the principal component matrix. Use the principal component matrix to reconstruct the original time series data matrix to obtain the feature data after the calibration data features are reconstructed.

8. A connector intelligent monitoring and control system according to claim 7, characterized in that: The training of the neural network model includes: Define recursive neural network as the basic structure of neural network model, which includes input layer, hidden layer and output layer; Take the time points on the unified time axis as time labels, add the time labels to the feature data, obtain the time series data set, use the time series data set as the training input of the neural network model, and use the control decision of the connector as the prediction output of the neural network model; Initialize the network parameters of the neural network model; the network parameters include the weight matrix from the input layer to the hidden layer, the cyclic weight matrix of the hidden layer, the bias vector of the hidden layer, the weight matrix from the hidden layer to the output layer, and the bias vector of the output layer; Define the loss function of the neural network model, Wherein, M is the number of data values ​​in the time series data set, and the data values ​​include current values ​​and voltage values; b is the index of the data value, α and β are weight parameters, and Jb is the actual electrical anomaly judgment value. is the electrical anomaly judgment value output by the model prediction, Eh is the actual environmental impact coefficient, Environmental impact coefficients for model prediction output; Get the sample set, set the time series data set as a subset, divide the sample set into multiple batches, pass each batch of the time series data set to the input layer in turn, update the network parameters according to the optimization algorithm, and stop when the loss function value is minimized and no longer changes, thus completing the training of the neural network model.

9. A connector intelligent monitoring and control system according to claim 8, characterized in that: The electrical abnormality determination value; Environmental impact factor, Among them, r1, r2, r3, r4, r5 and r6 are weight parameters; Tt is the real-time temperature value obtained within the cycle time; T1 is the maximum safe temperature; T2 is the minimum safe temperature; Sd is the real-time humidity value obtained within the cycle time; Dy is the real-time voltage value obtained within the cycle time; Dl is the real-time current value obtained within the cycle time; is the calibration voltage value; is the calibration current value.

10. A connector intelligent monitoring and control system according to claim 9, characterized in that: The proportion coefficient λ is calculated as, Wherein, v(b) represents the standard data value.

Images