Cable channel pressure change monitoring system based on sensor network

Abstract

The present invention relates to the field of cable fault detection technology, specifically to a cable channel pressure change monitoring system based on a sensor network, the system comprising: a pressure signal decomposition module collects multi-point pressure readings based on pressure signal data collected by the sensor network, and segmented filtering gradually divides the frequency of the signal. In the present invention, by collecting pressure data at multiple points, combining segmented filtering and hierarchical decomposition technology, the level of signal analysis is refined, the pressure characteristics of different frequencies are captured, and the comprehensiveness of monitoring is ensured. Energy entropy calculation is combined with wavelet decomposition method to achieve in-depth characterization of non-stationary characteristics of the signal and improve the detection sensitivity of abnormal changes. Time series prediction uses recursive analysis, combined with dynamic error window correction, to provide accurate prediction of future pressure changes. Mutation detection is based on dynamic thresholds and time point change rate judgment, and responds quickly to abnormal situations. Trend correction combines offset data and measurement data to effectively improve the prediction accuracy and correction ability of pressure change trends.

Description

Technical Field

[0001] The invention relates to the technical field of cable fault detection, and in particular to a cable channel pressure change monitoring system based on a sensor network. Background Art

[0002] The field of cable fault detection technology refers to the technology of using various monitoring and measurement methods to perform real-time or periodic detection of the status of cables in response to various faults that may occur during the operation of cables. The purpose of this field is to detect abnormalities or potential faults in cable operation in advance so that maintenance and repair can be carried out in time, thereby improving the stability and reliability of the power supply system. Typical cable faults include insulation aging, mechanical damage, temperature anomalies, pressure changes and other problems, and fault detection technology can use a variety of sensors and intelligent algorithms to diagnose these problems early to avoid major accidents or power outages caused by cable faults.

[0003] Among them, the cable channel pressure change monitoring system based on the sensor network refers to a system that monitors the pressure changes in the channel through multiple sensors arranged in the cable channel. Its main purpose is to monitor the pressure changes in the channel and find abnormal cable environment caused by water intrusion, foundation subsidence or other external factors, so as to give early warning and ensure the normal operation and safety of the cable.

[0004] Existing technologies lack effective decomposition of multi-frequency pressure signals in complex environments, and the acquisition method is single, resulting in insufficient sensitivity to pressure changes and difficulty in dealing with complex and diverse fault conditions. There is a lack of reasonable prediction of future pressure trends, and only post-processing can be relied upon, resulting in delayed warning. Detection of sudden anomalies is mostly based on fixed thresholds, ignoring the dynamic characteristics of pressure fluctuations, which is prone to misjudgment or omission, increasing the difficulty of maintenance and the risk of power outages, and affecting the stability and reliability of the power supply system. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a cable channel pressure change monitoring system based on a sensor network.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A cable channel pressure change monitoring system based on a sensor network includes:

[0007] The pressure signal decomposition module collects multi-point pressure readings based on the pressure signal data collected by the sensor network, performs segmented filtering to gradually divide the signal into different frequencies, selects a matching mother wavelet function for hierarchical decomposition, calls a multi-band signal set, and generates decomposed frequency band sub-signal values;

[0008] The energy entropy calculation module collects amplitude data for cumulative calculation based on the decomposed frequency band sub-signal values, compares energy ratios, determines wavelet energy entropy, calls entropy value sets for arrangement and marking, and obtains wavelet energy entropy eigenvalues;

[0009] The time series prediction module is based on the wavelet energy entropy eigenvalue, recursively analyzes multiple eigenvalues ​​step by step, generates a future prediction set in chronological order, outputs the future value of the corresponding entropy feature, obtains the energy entropy time series prediction value, sets the window to calculate the error item by item, and generates the time series prediction error value;

[0010] The mutation detection module calculates the error mean and variance based on the time series prediction error value, determines the dynamic threshold, marks the abnormal offset point, determines the trend by the entropy change rate of consecutive time points, obtains the offset situation, and generates the mutation offset determination result;

[0011] The trend correction module collects the offset data set to construct cumulative data based on the sudden change offset judgment result, eliminates the fluctuation data and generates a preliminary trend estimate, performs error correction in combination with the measurement data, and generates a pressure change trend correction result.

[0012] The decomposed frequency band sub-signal values ​​include low-frequency signal sub-bands, medium-frequency signal sub-bands, and high-frequency signal sub-bands; the wavelet energy entropy characteristic values ​​include low-frequency energy characteristics, medium-frequency energy characteristics, and high-frequency energy characteristics; the time series prediction error value includes single-point error, multi-point deviation, and trend error; the sudden shift determination result includes shift amplitude, shift direction, and shift identification; the pressure change trend correction result includes trend correction value, error correction coefficient, and trend change rate.

[0013] As a further solution of the present invention, the step of obtaining the decomposed frequency band sub-signal value is specifically as follows:

[0014] Collect multiple pressure readings from the sensor network, synchronize multiple reading times, obtain consistent time series data, and obtain a synchronized set of pressure readings;

[0015] Performing segmented filtering on the synchronous pressure reading set, applying low-pass and high-pass filters to the data in each time period, extracting signal features of differentiated frequency components, and obtaining a frequency feature distribution signal set;

[0016] A mother wavelet function matching the signal characteristics is selected from the frequency characteristic distribution signal set, and wavelet decomposition is performed step by step to extract the multi-band components of the signal, using the formula:

[0017] ;

[0018] Calculate sub-signal values ​​of multiple frequency bands to obtain a frequency band sub-signal set;

[0019] in, Representative The sub-signal value of the decomposed frequency band, Representative The amplification factor of the signal component, Represents the amplitude value of the signal, Indicates the signal used The mother wavelet function, Represents the weight parameter of high-pass filtering, which is used to adjust the criticality of high-frequency components. Represents the weight parameter of low-pass filtering, which is used to adjust the criticality of low-frequency components. It represents the frequency response value of the signal after high-pass filtering. It represents the frequency response value of the signal after low-pass filtering. represents the total number of signal components, The total number of signal data points representing the second part, indicating the number of samples of the second part;

[0020] The frequency band sub-signal sets are reorganized, sorted and merged according to the frequency characteristics of the signals to form a continuous signal spectrum, and the decomposed frequency band sub-signal values ​​are generated.

[0021] As a further solution of the present invention, the step of obtaining the wavelet energy entropy eigenvalue is specifically as follows:

[0022] Based on the decomposed frequency band sub-signal values, collecting and accumulating amplitude data of multiple frequency bands, calculating the total amplitude of each frequency band, and generating a cumulative amplitude data set;

[0023] Performing energy ratio analysis on the cumulative amplitude data set, calculating energy ratios of multiple frequency bands, determining energy distribution of the multiple frequency bands by numerical comparison, and obtaining energy ratio results;

[0024] The energy ratio result is used to calculate the wavelet energy entropy using the entropy calculation formula:

[0025] ;

[0026] Generate energy entropy characteristic data, where Indicates The wavelet energy entropy of the frequency band, Indicates The energy ratio of the frequency band, Indicates The logarithm of the frequency band energy ratio, Represents an adjustment parameter used to enhance the sensitivity of entropy to low-probability events. Indicates The inverse value of the band energy ratio, Represents the total number of signal frequency bands, indicating the number of all frequency bands involved in calculating energy entropy;

[0027] The energy entropy characteristic data are sorted and marked, multiple frequency bands are marked from high to low according to the entropy value, the frequency band with the most uneven energy distribution is identified, and the wavelet energy entropy characteristic value is obtained.

[0028] As a further solution of the present invention, the step of obtaining the time series prediction error value is specifically as follows:

[0029] Based on the wavelet energy entropy eigenvalue, the time series pattern is analyzed by a recursive method, the data of past time points are continuously calculated to predict future trends, and a feature prediction result is generated;

[0030] Using the feature prediction results, a future prediction set is constructed in chronological order, and a linear regression model is applied to predict the entropy feature values ​​at multiple future time points to obtain a future entropy feature value set;

[0031] For the future entropy feature value set, error calculation is performed and the time series prediction formula is applied:

[0032] ;

[0033] Calculate the prediction error and generate the energy entropy time series prediction value;

[0034] in, Represents the predicted value at the next time point, Representing the past The energy entropy eigenvalue at a time point, Representatives and past Energy entropy value at time point The regression weight coefficient represents the The contribution of the time point to the predicted value, represents the error adjustment factor used to adjust the effect of standard deviation, Represents the standard deviation of the entropy eigenvalue at the current time point, which is used to measure the volatility of the entropy eigenvalue. Represents a positive number that avoids the standard deviation from being zero, used to avoid division by zero errors. Represents the initial bias term of the time series, which is used to adjust the overall prediction level of the model. Represents the length of the previous data, which is the number of data points in the time series used for regression model training;

[0035] Using the energy entropy time series prediction value, a window is set to calculate the difference between the prediction value and the actual entropy value item by item, evaluate the accuracy of the prediction model, and adjust the model parameters to generate a time series prediction error value.

[0036] As a further solution of the present invention, the step of obtaining the mutation shift determination result is specifically as follows:

[0037] Starting from the time series prediction error value, calculating the mean and variance of the error to obtain statistical mean and variance results;

[0038] According to the statistical mean and variance results, a dynamic threshold is set to match the volatility of the current data and a dynamic threshold setting result is generated;

[0039] Based on the dynamic threshold setting result, the entropy change rate at consecutive time points is analyzed, the change rate is compared with the dynamic threshold, the trend change of the data is judged, and the trend is quantified using a mathematical model, using the formula:

[0040] ;

[0041] Determine whether the entropy changes at consecutive time points significantly exceed expectations and obtain trend analysis results;

[0042] in, Representing time point The trend assessment value of Representative The entropy value at a time point, represents the mean entropy value, is the adjustment coefficient, which is used to adjust the sensitivity of trend changes. represents the number of selected consecutive time points;

[0043] By using the trend analysis results, all abnormal deviation points are marked to indicate a sudden change or significant deviation of the entropy value, and the deviation of the abnormal deviation points is iteratively analyzed to generate a sudden deviation determination result.

[0044] As a further solution of the present invention, the step of obtaining the pressure change trend correction result is specifically:

[0045] Based on the sudden shift determination result, the shift data is collected and accumulated, the accumulated amount is calculated for each shift point, and the accumulated data is obtained through the accumulated result to obtain the accumulated shift data result;

[0046] Performing volatility analysis on the accumulated offset data results, by calculating the fluctuation range of each segment of data, judging and eliminating obvious short-term fluctuation data, retaining trend information, and generating a preliminary trend estimation result after removing fluctuations;

[0047] Combined with the preliminary trend estimation results after removing fluctuations, the measured data is used for error correction. Based on the difference between the actual measured data and the trend estimation, the formula is adopted:

[0048] ;

[0049] Calculate the corrected trend result to obtain the pressure change trend correction result;

[0050] in, Represents the corrected pressure change trend value, represents the preliminary estimate of the trend, Representative The actual measurement data of the measuring point, Represents the total number of measurement points, Representative Preliminary estimate of the trend corresponding to the measurement point, Indicates the deviation value of each measurement point.

[0051] Compared with the prior art, the advantages and positive effects of the present invention are:

[0052] In the present invention, by collecting pressure data at multiple points, combined with segmented filtering and hierarchical decomposition technology, the level of signal analysis is refined, the pressure characteristics of different frequencies are effectively captured, and the comprehensiveness of monitoring is ensured. Energy entropy calculation is combined with wavelet decomposition method to achieve in-depth characterization of non-stationary characteristics of the signal and improve the detection sensitivity of abnormal changes. Time series prediction uses recursive analysis, combined with dynamic error window correction, to provide accurate prediction of future pressure changes. Mutation detection is based on dynamic thresholds and time point change rate judgments to quickly respond to abnormal situations. Trend correction combines offset data with measurement data to effectively improve the prediction accuracy and correction ability of pressure change trends. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a system flow chart of the present invention;

[0054] Figure 2 A flow chart of the steps for obtaining the sub-signal values ​​of the decomposed frequency bands of the present invention;

[0055] Figure 3 This is a flow chart of the steps for obtaining the wavelet energy entropy eigenvalue of the present invention;

[0056] Figure 4 This is a flow chart of the steps for obtaining the time series prediction error value of the present invention;

[0057] Figure 5 A flowchart of the steps for obtaining the mutation shift determination result of the present invention;

[0058] Figure 6 This is a flow chart of the steps for obtaining the pressure change trend correction result of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0060] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0061] Example 1: Please refer to Figure 1 , the cable channel pressure change monitoring system based on the sensor network includes:

[0062] The pressure signal decomposition module collects multi-point pressure readings based on the pressure signal data collected by the sensor network, performs segmented filtering to gradually divide the signal into different frequencies, selects a matching mother wavelet function for hierarchical decomposition, calls a multi-band signal set, and generates decomposed frequency band sub-signal values;

[0063] The energy entropy calculation module collects amplitude data for cumulative calculation based on the decomposed frequency band sub-signal values, compares the energy ratio, determines the wavelet energy entropy, calls the entropy value set for arrangement and marking, and obtains the wavelet energy entropy eigenvalue;

[0064] The time series prediction module is based on the wavelet energy entropy eigenvalue, recursively analyzes multiple eigenvalues ​​step by step, generates future prediction sets in chronological order, outputs the future values ​​of the corresponding entropy features, obtains the energy entropy time series prediction values, sets the window to calculate the errors item by item, and generates the time series prediction error values;

[0065] The mutation detection module predicts the error value based on the time series, calculates the error mean and variance, determines the dynamic threshold, marks the abnormal offset point, determines the trend based on the entropy change rate of consecutive time points, obtains the offset situation, and generates the mutation offset judgment result;

[0066] The trend correction module collects the offset data set to construct cumulative data based on the sudden change offset judgment result, eliminates the fluctuating data and generates a preliminary trend estimate, combines the measured data for error correction, and generates a pressure change trend correction result.

[0067] The decomposed frequency band sub-signal values ​​include low-frequency signal sub-band, medium-frequency signal sub-band, and high-frequency signal sub-band. The wavelet energy entropy characteristic values ​​include low-frequency energy characteristics, medium-frequency energy characteristics, and high-frequency energy characteristics. The time series prediction error value includes single-point error, multi-point deviation, and trend error. The sudden change offset judgment result includes offset amplitude, offset direction, and offset mark. The pressure change trend correction result includes trend correction value, error correction coefficient, and trend change rate.

[0068] See also Figure 2 , the specific steps for obtaining the decomposed frequency band sub-signal value are as follows:

[0069] Collect multiple pressure readings from the sensor network, synchronize multiple reading times, obtain consistent time series data, and obtain a synchronized set of pressure readings;

[0070] When collecting multi-point pressure readings, it is necessary to ensure the time synchronization of the collected data. This requires that the data received from each node of the sensor network must have the same timestamp. By calibrating and time-synchronizing the sensors, the consistency of the data is guaranteed, which facilitates subsequent data analysis and processing. The time-synchronized set of pressure readings provides the basic data source for analysis.

[0071] Performing segmented filtering on the set of synchronized pressure readings, applying low-pass and high-pass filters to the data in each time period, extracting the signal characteristics of the differentiated frequency components, and obtaining a set of frequency characteristic distribution signals;

[0072] Segmented filtering of the synchronized data means dividing the continuous data stream into multiple time windows and applying low-pass and high-pass filters to each window. This method can effectively extract each frequency component from the original signal, especially in a noisy environment. The filter can help remove interference from non-target frequencies and improve the quality and reliability of signal processing. The frequency characteristic distribution signal set lays the foundation for the next step of waveform analysis and feature extraction.

[0073] Select the mother wavelet function that matches the signal characteristics from the frequency characteristic distribution signal set, perform wavelet decomposition step by step, and extract the multi-band components of the signal using the formula:

[0074] ;

[0075] Calculate sub-signal values ​​of multiple frequency bands to obtain a frequency band sub-signal set;

[0076] in, Representative The sub-signal value of the decomposed frequency band, Representative The amplification factor of the signal component, Represents the amplitude value of the signal, Indicates the signal used The mother wavelet function, Represents the weight parameter of high-pass filtering, which is used to adjust the criticality of high-frequency components. Represents the weight parameter of low-pass filtering, which is used to adjust the criticality of low-frequency components. It represents the frequency response value of the signal after high-pass filtering. It represents the frequency response value of the signal after low-pass filtering. represents the total number of signal components, The total number of signal data points representing the second part, indicating the number of samples of the second part;

[0077] formula:

[0078] ;

[0079] The benefit of the formula is that it combines the advantages of wavelet transform and filtering technology by adjusting the parameters , ,and To optimize the frequency domain decomposition and reconstruction of the signal, thereby improving the signal quality and frequency resolution.

[0080] Detailed explanation of the formula and the process of formula calculation and derivation:

[0081] set up , , , , , ,have:

[0082] ;

[0083] ;

[0084] ;

[0085] The results show that by proper parameter selection, the details and overall quality of the signal can be significantly optimized, especially in complex signal environments, where the formula can help extract key signal features and reduce information loss.

[0086] The frequency band sub-signal sets are reorganized, sorted and merged according to the frequency characteristics of the signals to form a continuous signal spectrum and generate decomposed frequency band sub-signal values.

[0087] The decomposed sub-signals are recombined, sorted and merged according to their frequency characteristics. This step is a key link in signal reconstruction. By sorting out the sub-signals of each frequency band, the approximate form of the original signal can be effectively reconstructed, which is crucial for subsequent signal analysis and application. For example, in sound recognition or seismic wave analysis, accurate signal reconstruction can greatly improve the accuracy and reliability of recognition. The generated decomposed frequency band sub-signal values ​​will be used in the next step of signal analysis and feature extraction.

[0088] See also Figure 3 , the specific steps for obtaining the wavelet energy entropy eigenvalue are:

[0089] Based on the decomposed frequency band sub-signal values, the amplitude data of multiple frequency bands are collected and accumulated, the total amplitude of each frequency band is calculated, and a cumulative amplitude data set is generated;

[0090] Based on the collection process of the decomposed frequency band sub-signal values, the amplitude data is extracted from the pressure data collected synchronously at each monitoring point, and the total amplitude of each frequency band is calculated and accumulated one by one. This operation involves mathematically adding the amplitude of the pressure signal to ensure the data integrity and accuracy of each frequency band. Through this method, the energy distribution of each frequency band is monitored and recorded in real time. This process requires high-precision data processing equipment to ensure the continuity and accuracy of data acquisition. At the same time, operators need to calibrate the equipment regularly to prevent equipment aging from affecting data accuracy. By analyzing the accumulated data, the health status of the structure can be monitored in real time or potential abnormalities can be warned to obtain a cumulative amplitude data set.

[0091] Perform energy ratio analysis on the cumulative amplitude data set, calculate the energy ratio of multiple frequency bands, determine the energy distribution of multiple frequency bands through numerical comparison, and obtain the energy ratio result;

[0092] The processing of the cumulative amplitude data set involves calculating the energy ratio within each frequency band. The energy ratio of each frequency band is obtained by comparing the cumulative amplitude of a single frequency band with the total energy. This calculation process requires precise numerical input to ensure the accuracy of the ratio. In this way, the energy contribution of different frequency bands can be compared to provide basic data for subsequent energy analysis. The energy ratio results obtained not only reflect the energy distribution of each frequency band, but are also very useful for identifying the main energy areas in the signal. This helps to give priority to energy-concentrated areas for analysis and processing when processing complex signals, thereby improving the efficiency and accuracy of signal processing.

[0093] Use the energy ratio results to calculate the wavelet energy entropy using the entropy calculation formula:

[0094] ;

[0095] Generate energy entropy characteristic data, where Indicates The wavelet energy entropy of the frequency band, Indicates The energy ratio of the frequency band, Indicates The logarithm of the frequency band energy ratio, Represents an adjustment parameter used to enhance the sensitivity of entropy to low-probability events. Indicates The inverse value of the band energy ratio, Represents the total number of signal frequency bands, indicating the number of all frequency bands involved in calculating energy entropy;

[0096] formula:

[0097] ;

[0098] The benefit of the formula is that by adding an inverse probability term, the sensitivity of the formula to low-probability events is enhanced, making the entropy calculation more sensitive when facing uneven data distribution. It is suitable for signal analysis of complex systems, especially when there are large differences in signal strength, and can better distinguish the information content.

[0099] Detailed explanation of the formula and the process of formula calculation and derivation:

[0100] Assume there are three frequency bands, the energy ratio of frequency band 1 is 0.2, frequency band 2 is 0.3, and frequency band 3 is 0.5. Adjust the parameters Set to 0.5. Then the entropy part of each frequency band is calculated as follows:

[0101] ;

[0102] ;

[0103] ;

[0104] Adding these values ​​together gives the overall wavelet energy entropy:

[0105] ;

[0106] The results show that the entropy value of the overall system is 6.3352, which reflects the energy distribution of the system in these three frequency bands. A higher entropy value means that the energy distribution of the system is more dispersed and there is no obvious energy concentration area. For signal processing, this may mean that it is necessary to further analyze the reasons for energy dispersion or adjust the signal processing strategy.

[0107] The energy entropy feature data are sorted and marked, multiple frequency bands are marked from high to low according to the entropy value, the frequency band with the most uneven energy distribution is identified, and the wavelet energy entropy feature value is obtained.

[0108] After completing the calculation of energy entropy, the steps of sorting and marking the energy entropy data involve using a sorting algorithm to arrange the entropy values ​​from high to low. This process requires not only accurate calculation of the data, but also consideration of data stability and processing speed during the calculation process. The sorted entropy value marking can help engineers quickly identify the frequency band with the most uneven energy distribution, which is very useful for system maintenance or fault diagnosis. In mechanical equipment monitoring, the energy distribution of the frequency band can indicate the wear of mechanical parts. The sorted and marked data provides an intuitive way to observe and analyze these phenomena, making maintenance and fault handling more efficient and targeted, and obtaining the wavelet energy entropy eigenvalue.

[0109] See also Figure 4 , the specific steps for obtaining the time series prediction error value are:

[0110] Based on the wavelet energy entropy eigenvalue, the time series pattern is analyzed through a recursive method, the data of past time points are continuously calculated to predict future trends and generate feature prediction results;

[0111] Through the recursive analysis method, combined with the existing wavelet energy entropy eigenvalue data, its time series characteristics are deeply explored. This process involves analyzing the previous data through a recursive algorithm to predict future trends. First, the entropy value of each time point is regarded as an independent feature, and then statistical methods such as autoregressive models are applied to estimate the future entropy value. By comparing the previous data of the same period, a prediction model is gradually constructed. The model can predict possible future trends based on the previous changes in entropy values. This process ensures that each step of the calculation is based on actual data and statistical principles. The generated time series feature prediction results not only provide an intuitive preview of future trends, but also provide a scientific basis for further data analysis and decision making.

[0112] Using the feature prediction results, a future prediction set is constructed in chronological order, and a linear regression model is applied to predict the entropy feature values ​​at multiple future time points to obtain a future entropy feature value set;

[0113] According to the prediction results obtained from time series analysis, the linear regression model is used to construct the future prediction set in chronological order. In this process, the entropy characteristic value of each prediction is arranged in chronological order in detail. The generation of each data point is based on the recursive analysis result of the previous step. The mathematical modeling method is used to achieve this prediction. Specifically, the least squares method is used to estimate the parameters of the regression line to ensure that the prediction results of the model not only reflect the actual change trend of the data, but also minimize the prediction error. Through this method, the entropy characteristic value of each time point in the future can be more accurately displayed, thereby providing data support for subsequent decision-making.

[0114] For the future entropy feature value set, the error is calculated and the time series prediction formula is applied:

[0115] ;

[0116] Calculate the prediction error and generate the energy entropy time series prediction value;

[0117] in, Represents the predicted value at the next time point, Representing the past The energy entropy eigenvalue at a time point, Representatives and past Energy entropy value at time point The regression weight coefficient represents the The contribution of the time point to the predicted value, represents the error adjustment factor used to adjust the effect of standard deviation, Represents the standard deviation of the entropy eigenvalue at the current time point, which is used to measure the volatility of the entropy eigenvalue. Represents a positive number that avoids the standard deviation from being zero, used to avoid division by zero errors. Represents the initial bias term of the time series, which is used to adjust the overall prediction level of the model. Represents the length of the previous data, which is the number of data points in the time series used for regression model training;

[0118] formula:

[0119] ;

[0120] The benefit of the formula is that it makes the forecast more accurate by considering the changing trend of previous data and the impact of random fluctuations, and is suitable for complex time series data analysis.

[0121] Detailed explanation of the formula and the process of formula calculation and derivation:

[0122] First set , , , .assumed The previous data is {0.3, 0.45, 0.65}, where , that is, considering the previous data of three time points. Calculate the predicted value at the next time point according to the formula:

[0123] ;

[0124] The calculation steps are as follows:

[0125] Calculate the weighted sum of the previous data:

[0126] ;

[0127] 2. Calculate the impact of random fluctuations:

[0128] ;

[0129] 3. Sum all parts:

[0130] ;

[0131] The result shows that the predicted energy entropy at the next time point is 1.15, reflecting the possible growth trend of entropy values ​​in the future, which is consistent with the method of obtaining the time series prediction value in the step results.

[0132] Using the energy entropy time series forecast value, a window is set to calculate the difference between the forecast value and the actual entropy value item by item, evaluate the accuracy of the forecast model, and adjust the model parameters to generate the time series forecast error value.

[0133] By setting a specific calculation window, the difference between the time series prediction value and the actual entropy value is compared item by item. This process involves detailed error analysis, calculating the difference between the prediction result and the actual observation value at each step, and then evaluating the accuracy of the entire prediction model. Through error analysis at each time point, the parameters of the prediction model can be dynamically adjusted to adapt to new trends or fluctuations that may appear in the data. This detailed description clearly shows how to optimize the model by continuously monitoring the error, making the model closer to the behavior of the actual data, thereby improving the accuracy of the prediction, ensuring that the generated time series prediction error value is minimized, and providing reliable data support for future decision-making.

[0134] See also Figure 5 , the specific steps for obtaining the mutation shift determination result are:

[0135] Starting from the time series forecast error value, calculate the mean and variance of the error to obtain the statistical mean and variance results;

[0136] Starting from the time series prediction error value, the mean and variance of these errors are accurately calculated. These two statistical parameters provide basic reference values ​​for subsequent analysis. The calculation of the mean involves adding up all the error values ​​and dividing them by the total number of errors, while the calculation of the variance is the mean of the sum of the squares of each error value minus the mean. This method can effectively measure the fluctuation size and dispersion of the error value, which is particularly critical for the identification of outliers, because high variance indicates large prediction errors, while low variance indicates good stability of the model prediction. The obtained statistical mean and variance results not only help to understand the past prediction performance, but also provide a basis for adjusting the parameters of the future prediction model, so as to reduce errors in future predictions and improve the accuracy of predictions.

[0137] According to the statistical mean and variance results, dynamic thresholds are set to match the volatility of current data and generate dynamic threshold setting results;

[0138] A dynamic threshold is set based on the mean and variance. The calculation of this threshold is based on the error statistics. The method of dynamically adjusting the threshold takes into account the average level and range of error, allowing the system to adaptively adjust its anomaly detection sensitivity in a fluctuating data environment. By setting the threshold higher than the mean plus twice the variance, those anomalies that truly deviate from expectations can be effectively marked, which not only reduces the possibility of false alarms, but also improves the accuracy of anomaly detection. The dynamic threshold setting results make subsequent data processing more accurate, ensuring the quality and reliability of data analysis.

[0139] Based on the dynamic threshold setting results, the entropy change rate at consecutive time points is analyzed, the change rate is compared with the dynamic threshold, the trend change of the data is judged, and the trend is quantified using a mathematical model. The formula is:

[0140] ;

[0141] Determine whether the entropy changes at consecutive time points significantly exceed expectations and obtain trend analysis results;

[0142] in, Representing time point The trend assessment value of Representative The entropy value at a time point, represents the mean entropy value, is the adjustment coefficient, which is used to adjust the sensitivity of trend changes. represents the number of selected consecutive time points;

[0143] formula:

[0144] ;

[0145] The benefit of the formula is that it provides a method to quantify the trend of time series data. By calculating the weighted average of continuous entropy changes, the k value can be flexibly adjusted to adapt to different data characteristics, making the model's sensitivity adjustment to trends more detailed and adaptable.

[0146] Detailed explanation of the formula and the process of formula calculation and derivation:

[0147] Set n=3, , , k=1.5. The calculation process is as follows:

[0148] ;

[0149] ;

[0150] The result shows that under the given parameter settings, the trend evaluation value is 0, indicating that the entropy change at the current time point does not deviate significantly from the average level and the trend remains stable.

[0151] Using the trend analysis results, all abnormal deviation points are marked to indicate a sudden change or significant deviation in the entropy value. The deviation of the abnormal deviation points is analyzed iteratively to generate a sudden deviation determination result.

[0152] Using the trend analysis results, all abnormal deviation points are marked. These points indicate possible entropy mutations or major deviations. The specific deviations of these points are further analyzed, and the mutation deviation judgment results are generated by combining the previous steps. By setting specific marking standards, such as deviations from the mean exceeding three times the standard deviation, it is marked as abnormal. This method can not only accurately identify abnormal points in a statistical sense, but also adjust the sensitivity according to actual conditions. For example, when the data volatility is large, the multiple of the standard deviation can be appropriately reduced to avoid too many false alarms. This processing strategy ensures that each marked point has a clear statistical basis, ensures the objectivity and reliability of the processing results, and enables the mutation deviation judgment results to truly reflect the dynamic changes of the data, providing a solid foundation for subsequent data analysis and decision-making.

[0153] See also Figure 6 , the specific steps for obtaining the pressure change trend correction result are:

[0154] Based on the sudden shift determination result, the shift data is collected and accumulated, the accumulated amount is calculated for each shift point, and the accumulated data is obtained through the accumulated results to obtain the accumulated shift data result;

[0155] Based on the results of mutation shift judgment, we first collect specific shift data and perform accumulation processing on these data. Through accumulation processing, we can effectively eliminate short-term fluctuations in the data, making the trend more obvious. This process mainly relies on the combined use of numerical accumulation algorithm and data smoothing technology. The accumulation algorithm can quickly superimpose data points, while data smoothing technology helps filter out non-trend fluctuation noise. In this way, we can get a clear preliminary estimate of the trend, which provides a basis for subsequent error correction.

[0156] Perform volatility analysis on the accumulated offset data results. By calculating the fluctuation range of each segment of data, we can determine and eliminate obvious short-term fluctuation data, retain trend information, and generate preliminary trend estimation results after removing fluctuations.

[0157] Combining the preliminary trend estimate with the actual measured data for error correction is a critical step that directly affects the accuracy and reliability of the final trend. In the error correction process, it is first necessary to calculate the deviation between the trend value and the actual measured value of each data point, and then average these deviation values ​​to reduce the impact of random errors. In this way, the preliminary estimated trend can be adjusted to make it closer to the actual data changes. This process involves not only basic data processing skills, but also an in-depth understanding of data quality and accuracy to ensure that the trend correction results are both accurate and representative.

[0158] Combined with the preliminary trend estimation results after removing fluctuations, the measured data is used for error correction. Based on the difference between the actual measured data and the trend estimation, the formula is used:

[0159] ;

[0160] Calculate the corrected trend result to obtain the pressure change trend correction result;

[0161] in, Represents the corrected pressure change trend value, represents the preliminary estimate of the trend, Representative The actual measurement data of the measuring point, Represents the total number of measurement points, Representative Preliminary estimate of the trend corresponding to the measurement point, Indicates the deviation value of each measurement point.

[0162] formula:

[0163] ;

[0164] The benefit of the formula is that it can be dynamically adjusted directly based on the deviation between actual measured data and preliminary estimates, thereby accurately correcting trend forecasts. This dynamic adjustment method can respond to data changes in real time and improve the adaptability and accuracy of the forecast.

[0165] Detailed explanation of the formula and the process of formula calculation and derivation:

[0166] Assume that the initial trend estimate is The actual measured value is , calculate the difference between each measurement and its corresponding estimate , summing and averaging all the differences, we get , and then adding this mean deviation back to the initial estimate to get the corrected trend ,if , ,but:

[0167] ;

[0168] final:

[0169] ;

[0170] The results show that by considering the deviation between actual measurements and preliminary estimates, trend forecasts can be adjusted more accurately to make them closer to actual data changes, providing more reliable support for decision-making.

[0171] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A cable channel pressure change monitoring system based on a sensor network, characterized in that: The system comprises: The pressure signal decomposition module collects multi-point pressure readings based on the pressure signal data collected by the sensor network, performs segmented filtering to gradually divide the signal into different frequencies, selects a matching mother wavelet function for hierarchical decomposition, calls a multi-band signal set, and generates decomposed frequency band sub-signal values; The step of obtaining the decomposed frequency band sub-signal value is specifically as follows: Collect multiple pressure readings from the sensor network, synchronize multiple reading times, obtain consistent time series data, and obtain a synchronized set of pressure readings; Performing segmented filtering on the set of synchronous pressure readings, applying low-pass and high-pass filters to the data in each time period, extracting signal features of differentiated frequency components, and obtaining a set of frequency feature distribution signals; A mother wavelet function matching the signal characteristics is selected from the frequency characteristic distribution signal set, and wavelet decomposition is performed step by step to extract the multi-band components of the signal, using the formula: ； Calculate sub-signal values ​​of multiple frequency bands to obtain a frequency band sub-signal set; in, Representative The sub-signal value of the decomposed frequency band, Representative The amplification factor of the signal component, Represents the amplitude value of the signal, Indicates the signal used The mother wavelet function, Represents the weight parameter of high-pass filtering, which is used to adjust the criticality of high-frequency components. Represents the weight parameter of low-pass filtering, which is used to adjust the criticality of low-frequency components. It represents the frequency response value of the signal after high-pass filtering. It represents the frequency response value of the signal after low-pass filtering. represents the total number of signal components, The total number of signal data points representing the second part, indicating the number of samples of the second part; Recombining the frequency band sub-signal set, sorting and merging them according to the frequency characteristics of the signal to form a continuous signal spectrum, and generating decomposed frequency band sub-signal values; The energy entropy calculation module collects amplitude data for cumulative calculation based on the decomposed frequency band sub-signal values, compares energy ratios, determines wavelet energy entropy, calls entropy value sets for arrangement and marking, and obtains wavelet energy entropy eigenvalues; The time series prediction module is based on the wavelet energy entropy eigenvalue, recursively analyzes multiple eigenvalues ​​step by step, generates a future prediction set in chronological order, outputs the future value of the corresponding entropy feature, obtains the energy entropy time series prediction value, sets the window to calculate the error item by item, and generates the time series prediction error value; The mutation detection module calculates the error mean and variance based on the time series prediction error value, determines the dynamic threshold, marks the abnormal offset point, determines the trend by the entropy change rate of consecutive time points, obtains the offset situation, and generates the mutation offset determination result; The trend correction module collects the offset data set to construct cumulative data based on the sudden change offset determination result, eliminates the fluctuation data and generates a preliminary trend estimate, performs error correction in combination with the measurement data, and generates a pressure change trend correction result; The steps for obtaining the pressure change trend correction result are specifically as follows: Based on the sudden shift determination result, the shift data is collected and accumulated, the accumulated amount is calculated for each shift point, and the accumulated data is obtained through the accumulated result to obtain the accumulated shift data result; Performing volatility analysis on the accumulated offset data results, by calculating the fluctuation range of each segment of data, judging and eliminating obvious short-term fluctuation data, retaining trend information, and generating a preliminary trend estimation result after removing fluctuations; Combined with the preliminary trend estimation results after removing fluctuations, the measured data is used for error correction. Based on the difference between the actual measured data and the trend estimation, the formula is adopted: ； Calculate the corrected trend result to obtain the pressure change trend correction result; in, Represents the corrected pressure change trend value, represents the preliminary estimate of the trend, Representative The actual measurement data of the measuring point, Represents the total number of measurement points, Representative Preliminary estimate of the trend corresponding to the measurement point, Indicates the deviation value of each measurement point.

2. The cable channel pressure change monitoring system based on sensor network according to claim 1 is characterized in that: The decomposed frequency band sub-signal values ​​include low-frequency signal sub-bands, medium-frequency signal sub-bands, and high-frequency signal sub-bands; the wavelet energy entropy characteristic values ​​include low-frequency energy characteristics, medium-frequency energy characteristics, and high-frequency energy characteristics; the time series prediction error value includes single-point error, multi-point deviation, and trend error; the sudden shift determination result includes shift amplitude, shift direction, and shift identification; the pressure change trend correction result includes trend correction value, error correction coefficient, and trend change rate.

3. The cable channel pressure change monitoring system based on sensor network according to claim 1 is characterized in that: The steps for obtaining the wavelet energy entropy eigenvalue are specifically as follows: Based on the decomposed frequency band sub-signal values, collecting and accumulating amplitude data of multiple frequency bands, calculating the total amplitude of each frequency band, and generating a cumulative amplitude data set; Performing energy ratio analysis on the cumulative amplitude data set, calculating energy ratios of multiple frequency bands, determining energy distribution of the multiple frequency bands by numerical comparison, and obtaining energy ratio results; The energy ratio result is used to calculate the wavelet energy entropy using the entropy calculation formula: ； Generate energy entropy characteristic data, in, Indicates The wavelet energy entropy of the frequency band, Indicates The energy ratio of the frequency band, Indicates The logarithm of the band energy ratio, Represents an adjustment parameter used to enhance the sensitivity of entropy to low-probability events. Indicates The inverse value of the band energy ratio, Represents the total number of signal frequency bands, indicating the number of all frequency bands involved in calculating energy entropy; The energy entropy characteristic data are sorted and marked, multiple frequency bands are marked from high to low according to the entropy value, the frequency band with the most uneven energy distribution is identified, and the wavelet energy entropy characteristic value is obtained.

4. The cable channel pressure change monitoring system based on sensor network according to claim 3 is characterized in that: The steps for obtaining the time series prediction error value are specifically as follows: Based on the wavelet energy entropy eigenvalue, the time series pattern is analyzed by a recursive method, the data of past time points are continuously calculated to predict future trends, and a feature prediction result is generated; Using the feature prediction results, a future prediction set is constructed in chronological order, and a linear regression model is applied to predict the entropy feature values ​​at multiple future time points to obtain a future entropy feature value set; For the future entropy feature value set, error calculation is performed and the time series prediction formula is applied: ； Calculate the prediction error and generate the energy entropy time series prediction value; in, Represents the predicted value at the next time point, Representing the past The energy entropy eigenvalue at a time point, Representatives and past Energy entropy value at a time point The regression weight coefficient between The contribution of the time point to the predicted value, represents the error adjustment factor used to adjust the effect of standard deviation, Represents the standard deviation of the entropy eigenvalue at the current time point, which is used to measure the volatility of the entropy eigenvalue. Represents a positive number that avoids the standard deviation from being zero, used to avoid division by zero errors. Represents the initial bias term of the time series, which is used to adjust the overall prediction level of the model. Represents the length of the previous data, which is the number of data points in the time series used for regression model training; Using the energy entropy time series prediction value, a window is set to calculate the difference between the prediction value and the actual entropy value item by item, evaluate the accuracy of the prediction model, and adjust the model parameters to generate a time series prediction error value.

5. The cable channel pressure change monitoring system based on sensor network according to claim 4 is characterized in that: The steps for obtaining the mutation shift determination result are specifically as follows: Starting from the time series prediction error value, calculating the mean and variance of the error to obtain statistical mean and variance results; According to the statistical mean and variance results, a dynamic threshold is set to match the volatility of the current data and a dynamic threshold setting result is generated; Based on the dynamic threshold setting result, the entropy change rate at consecutive time points is analyzed, the change rate is compared with the dynamic threshold, the trend change of the data is judged, and the trend is quantified using a mathematical model, using the formula: ； Determine whether the entropy changes at consecutive time points significantly exceed expectations and obtain trend analysis results; in, Representing time point The trend assessment value of Representative The entropy value at a time point, represents the mean entropy value, is the adjustment coefficient, which is used to adjust the sensitivity of trend changes. represents the number of selected consecutive time points; By using the trend analysis results, all abnormal deviation points are marked to indicate a sudden change or significant deviation of the entropy value, and the deviation of the abnormal deviation points is iteratively analyzed to generate a sudden deviation determination result.

Images