A power equipment data intelligent analysis method and system based on 5G communication

Abstract

The application discloses a power equipment data intelligent analysis method and system based on 5G communication and relates to the technical field of power data analysis, solves the technical problems that the traditional scheme only uses static statistics or single time sequence characteristics, influences the accuracy of fault analysis results, and all data processing depends on the cloud, resulting in that the response time of time delay sensitive application is too long, constructs an abnormality evaluation model based on time sequence data characteristic coefficient formula through a data transfer node, can complete real-time analysis locally, can quickly trigger an early warning signal to corresponding sub-regional equipment when detecting a fault data transfer node, and reduces cloud computing tasks, fuses forecast data provided by a meteorological department and power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data, reduces original high-frequency data to a low-frequency sequence through coarse-grained downsampling, and forms a high-dimensional multi-scale feature space through time-frequency-space multidimensional fusion.

Description

Technical Field

[0001] This invention belongs to the field of power data analysis, specifically a method and system for intelligent analysis of power equipment data based on 5G communication. Background Technology

[0002] Power equipment is the core carrier for the safe and stable operation of the power system, encompassing power generation equipment, transmission equipment, substation equipment, and distribution equipment. With the continuous growth of social electricity demand, the power system is showing a development trend of "high voltage, large capacity, and wide coverage," with a surge in the number of devices and increasingly complex operating conditions. Traditional power equipment operation and maintenance relies on a "periodic inspection + manual diagnosis" model, which has three major pain points: First, poor timeliness, with manual inspection cycles typically lasting 1-3 months, making it impossible to capture sudden equipment failures in real time; second, limited coverage, with remote mountainous areas and transmission lines crossing rivers and seas making high-frequency inspections difficult, easily creating monitoring blind spots; and third, low data utilization, with manually recorded parameters such as temperature and voltage mostly stored in paper archives, making long-term trend analysis and intelligent early warning impossible. Against this backdrop, "condition-based maintenance" has become the core development direction of the power industry—by collecting real-time equipment operating data and analyzing equipment operating status, an intelligent data analysis solution for power equipment is needed.

[0003] In traditional power equipment data analysis solutions, sample entropy or approximate entropy only analyzes a single variable and cannot quantify the coupled impact of meteorological disasters and equipment status. Traditional models only use static statistics or single time-series features, which affects the accuracy of fault analysis results. At the same time, in traditional architectures, all data processing relies on the cloud, resulting in excessively long response times for latency-sensitive applications. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method and system for intelligent analysis of power equipment data based on 5G communication, which is used to solve the technical problems that traditional solutions only use static statistics or single time-series features, affecting the accuracy of fault analysis results, and that all data processing depends on the cloud, resulting in excessively long response times for latency-sensitive applications.

[0005] To address the aforementioned problems, the first aspect of this invention provides a method and system for intelligent analysis of power equipment data based on 5G communication, comprising the following steps:

[0006] Based on the type of power equipment, the power equipment is divided into several subcategories. Operational data of the power equipment in each subcategory is collected, and geographical location data and local meteorological data of the power equipment in each subcategory are obtained.

[0007] Based on the distribution of power equipment, the entire power area is divided into multiple sub-areas. Each sub-area is equipped with a 5G communication base station to transmit operational data and meteorological data to a data relay node. The data relay node then regroups the acquired data according to the type of power equipment and sends the grouped data to the cloud.

[0008] The data transfer node analyzes the characteristic coefficients according to the characteristic index formula based on the time series data of different monitoring indicators of the operation data, constructs the operation anomaly assessment model, and sends early warning signals to the power equipment with operation anomalies in the corresponding sub-region.

[0009] The cloud-based analysis system integrates forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data; for different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on the fused meteorological data;

[0010] The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales.

[0011] Based on coarse-grained sequences at different time scales, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed to calculate the improved MMSE value of power equipment.

[0012] Based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales, fault feature vectors are constructed; by analyzing the distribution characteristics of fault feature vectors under different fault types, power equipment faults are classified.

[0013] Optionally, in one example of the above aspects, the subcategories include: transmission lines, power generation equipment, transformer equipment, and distribution equipment; the operating data of the power equipment includes: current, voltage, power, temperature, humidity, and amplitude; the meteorological data types include: temperature, humidity, wind speed, precipitation, and lightning density, and the meteorological data includes meteorological data collected by meteorological monitoring sensors around the power equipment, as well as forecast data provided by the meteorological department;

[0014] Optionally, in one example of the above aspects, characteristic coefficient analysis is performed based on time-series data of different monitoring indicators of operational data, according to the characteristic indicator formula, including the following steps:

[0015] The operational data of each group of power equipment is standardized. From the time-series data of each monitoring indicator of the standardized power equipment operational data, for any monitoring indicator, a time period [t] is set for the operational data monitoring indicator. a ,t bThe time interval is divided into different intervals, with corresponding time interval numbers (1, 2, ..., n). The standardized mean of the data within each time interval is taken as the feature time series data: (x1′, x2′, ..., x n ′); Analyze trend characteristic coefficients, fluctuation characteristic coefficients, and abrupt change characteristic coefficients;

[0016] Optionally, in one example of the above aspects, building and running an anomaly assessment model includes the following steps:

[0017] Obtain historical data and statistically analyze the trend characteristic coefficient, fluctuation characteristic coefficient, and sudden change characteristic coefficient of the equipment;

[0018] Based on the historical operating status of the equipment, the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that is operating normally and is not in the preset time interval before the failure are marked as normal operation; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that has failed are marked as failure; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment in the preset time interval before the failure are marked as potential failure.

[0019] LSTM models are trained for different monitoring indicators to identify the real-time operating status of electrical equipment.

[0020] Optionally, in one example of the above aspects, the forecast data provided by the meteorological department is fused with the power grid meteorological monitoring data collected by meteorological monitoring sensors around the power equipment to obtain fused meteorological data, including the following steps:

[0021] Let the meteorological forecast data be X. m The meteorological monitoring data for the power grid is X. g Let w be the weight of the meteorological forecast data. m The weight of the power grid meteorological monitoring data is w g And w m +w g =1;

[0022] Weighted fusion of meteorological data Among them, the meteorological department's forecast data parameter w m =exp(-(t) m -t g ) / t0), t m t represents the time difference between the latest update time of meteorological data and the current time, as predicted by the meteorological department. g t0 represents the time difference between the latest update time of the meteorological data in the power grid meteorological monitoring data and the current time, and t0 represents the basic time interval for updating the meteorological data in the meteorological department's forecast data.

[0023] Statistical analysis of meteorological forecast data parameters w based on historical data m The maximum value of ′ for w m After normalization, the weights of the meteorological forecast data are obtained as w. m .

[0024] Optionally, in one example of the above aspects, key meteorological influencing factors are selected for different equipment types, and a disaster impact quantification index is constructed based on fused meteorological data, including the following steps:

[0025] Based on fused meteorological data, key meteorological influencing factors were selected for different equipment types:

[0026] For transmission lines, the key meteorological influencing factors are set as wind speed, temperature, humidity, and lightning density;

[0027] For power generation equipment, the key meteorological influencing factors are set as temperature, precipitation, and lightning density;

[0028] For power equipment and distribution equipment, the key meteorological influencing factors are set as temperature, humidity and precipitation;

[0029] The formula for calculating the impact index of key meteorological influencing factors is set as follows:

[0030]

[0031] Among them, Ih is the impact index of key meteorological influencing factors, and th j Let tj be the duration of the j-th fused data for the key meteorological impact factors, th0 be the basic update time interval for the key meteorological impact factors, and xh be the duration of the data for the j-th fused data for the key meteorological impact factors. j Let xh be the value of the j-th fused data for the key meteorological influencing factors. min xh represents the minimum historical value of the key meteorological influencing factors. max is the maximum value of the historical data of the key meteorological impact factor, j∈(1,2,…,m), and m is the total number of updates of the fused data of the key meteorological impact factor;

[0032] The formula for the disaster impact quantification index is set as a weighted average formula for the impact index of key meteorological impact factors corresponding to different equipment types.

[0033] Optionally, in one example of the above aspects, a multivariate time series is constructed by combining the disaster impact quantification index, operational anomaly assessment data, and power equipment operation data; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales, including the following steps:

[0034] For each variable dimension of the multivariate time series consisting of disaster impact quantification index, operational anomaly assessment data, and power equipment operation data, a coarse-grained sequence is generated according to a preset time scale factor.

[0035] The coarse-grained sequence is obtained under the preset time scale factor s of the variable dimensions: Y(s)(g)={Y1(s)(g),Y2(s)(g),…,Y r (s)(g)};

[0036] Based on the coarse-grained sequence Y(s)(g) of the g-th variable dimension and the coarse-grained sequence Y(s)(h)={Y1(s)(h),Y2(s)(h),…,Y r By calculating the similarity matrix between the g-th and h-th variable dimensions (s)(h)}, we obtain the similarity matrix D among the multiple variables. s ;

[0037] Elements in the matrix , where Y a (s)(g) represents the a-th coarse-grained sequence parameter of the g-th variable dimension under the time scale factor s, Y a (s)(h) represents the a-th coarse-grained sequence parameter of the h-th variable dimension under the time scale factor s, where a∈(1,2,…,r), and r is the total number of coarse-grained sequence parameters of the g-th variable dimension under the time scale factor s.

[0038] Optionally, in one example of the above aspects, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed based on coarse-grained sequences at different time scales to calculate the improved MMSE value of power equipment, including the following steps:

[0039] For the similarity matrix D between multivariates of the coarse-grained sequence at each time scale s, s Based on the adaptive similarity tolerance parameter rg, the matching probability C of dimension ds at time scale s is calculated. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r);

[0040] Among them, C s (ds)(r)= ; D is a similarity matrix of dimension ds among multiple variables. s In the middle The element, R s For similarity matrix D s The adaptive similarity tolerance parameter, Let the Heavgsgdse step function be the sequence length, and N0 be the sequence length; similarly, calculate C.s (ds+1)(r);

[0041] An improved multivariate multiscale sample entropy (MMSE) calculation model is constructed, based on the matching probability C with dimension ds. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r) is used to calculate the multivariate multiscale sample entropy, which serves as the improved MMSE value for power equipment: MMSE(r,s,R) s ,N0)=-ln(C s (ds+1)(r) / C s (ds)(r)); where r is the total number of coarse-grained sequence parameters of the g-th variable dimension under the time scale factor s.

[0042] Optionally, in one example of the above aspects, power equipment fault classification is performed by analyzing the distribution characteristics of fault feature vectors under different fault types, including the following steps:

[0043] The improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales are combined with trend characteristic coefficients, fluctuation characteristic coefficients and abrupt change characteristic coefficients to form a high-dimensional fault feature vector;

[0044] The ReliefF algorithm is used to remove redundant features, and the distribution characteristics of high-dimensional fault feature vectors under different fault types are analyzed by box plots or t-SNE dimensionality reduction techniques.

[0045] The LSTM (Long Short-Term Memory) network model is trained by analyzing the distribution characteristics of high-dimensional fault feature vectors under different fault types in historical data. The spatiotemporal correlation of the distribution characteristics is learned, and the trained LSTM network model is used to classify power equipment faults.

[0046] According to another aspect of this disclosure, a power equipment data intelligent analysis system based on 5G communication is provided, comprising: a data acquisition layer, a communication transmission layer, and a cloud analysis layer. The system employs the aforementioned power equipment data intelligent analysis method based on 5G communication to achieve intelligent analysis of power equipment data.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This invention constructs an anomaly assessment model based on the time-series data feature coefficient formula through data relay nodes. It can perform real-time analysis locally. When a faulty data relay node is detected, an early warning signal can be quickly triggered to the corresponding sub-area device, avoiding the expansion of the fault due to cloud computing delay. The data processing relay nodes are optimized by grouping them according to device type to classify the data. Dedicated analysis algorithms can be applied to different device types to improve the accuracy of feature extraction.

[0049] This invention fuses forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data, which can eliminate the bias of a single data source; by coarsening the original high-frequency data into low-frequency sequences, sensor noise is effectively suppressed while retaining fault characteristics. Coarsened sequences at different scales can reflect different stages of fault development, providing more comprehensive feature input for the model; and through time-frequency-spatial multi-dimensional fusion, a high-dimensional distinguishable feature space is formed. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0052] Figure 2 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0053] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please see Figures 1-2 The first aspect of this invention provides a method and system for intelligent analysis of power equipment data based on 5G communication, comprising the following steps:

[0055] Based on the type of power equipment, the power equipment is divided into several subcategories. Operational data of the power equipment in each subcategory is collected, and geographical location data and local meteorological data of the power equipment in each subcategory are obtained.

[0056] Based on the distribution of power equipment, the entire power area is divided into multiple sub-areas. Each sub-area is equipped with a 5G communication base station to transmit operational data and meteorological data to a data relay node. The data relay node then regroups the acquired data according to the type of power equipment and sends the grouped data to the cloud.

[0057] The data transfer node analyzes the characteristic coefficients according to the characteristic index formula based on the time series data of different monitoring indicators of the operation data, constructs the operation anomaly assessment model, and sends early warning signals to the power equipment with operation anomalies in the corresponding sub-region.

[0058] The cloud-based analysis system integrates forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data; for different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on the fused meteorological data;

[0059] The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales.

[0060] Based on coarse-grained sequences at different time scales, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed to calculate the improved MMSE value of power equipment. For each coarse-grained sequence at each time scale, the similarity matrix between the multivariates is calculated. Based on the adaptive similarity tolerance parameter, the matching probability at the time scale with dimensions d and d+1 is calculated.

[0061] Based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales, fault feature vectors are constructed; by analyzing the distribution characteristics of fault feature vectors under different fault types, power equipment faults are classified.

[0062] In one embodiment of the present invention, the subcategories include: transmission lines, power generation equipment, transformer equipment, and distribution equipment; the operating data of the power equipment includes: current, voltage, power, temperature, humidity, and amplitude; the meteorological data includes: temperature, humidity, wind speed, precipitation, and lightning density, and the meteorological data includes meteorological data collected by meteorological monitoring sensors around the power equipment, as well as forecast data provided by the meteorological department;

[0063] The standardization of the operating data for each group of power equipment includes the following steps:

[0064] Standardize the operating data of different types of power equipment (such as voltage U, current I, and temperature T);

[0065] Let the rated value of a certain monitoring indicator be Se, and the noise-reduced data at time tk be x. tk The standardized running data is x tk ′, then: x tk ′=(x tk -Se) / Se×100%, this formula converts the data into a percentage deviation relative to the nominal value.

[0066] When x tk When x' = 0, the data fully meets the rated requirements; when x' = 0, the data fully meets the rated requirements. tk When x > 0, the data is higher than the rated value; when x > 0, the data is higher than the rated value. tk When ′<0, the data is lower than the rated value, which makes it easier to intuitively reflect the degree of deviation between the data and the normal operating status of the equipment.

[0067] In one embodiment of the present invention, based on time-series data of different monitoring indicators of operational data, characteristic coefficient analysis is performed according to the characteristic indicator formula, including the following steps:

[0068] The operating data of each group of power equipment is standardized, and the trend characteristic coefficient, fluctuation characteristic coefficient and sudden change characteristic coefficient are analyzed from the time series data of each monitoring indicator of the standardized power equipment operating data.

[0069] For any monitoring indicator, set the time period [t] for the operational data monitoring indicator. a ,t b The time interval is divided into different intervals, with corresponding time interval numbers (1, 2, ..., n). The standardized mean of the data within each time interval is taken as the feature time series data: (x1′, x2′, ..., x n ′);

[0070] Based on the trend characteristics and the characteristic time series data, the time interval [t] is calculated. a ,t b The linear trend slope characteristic coefficient ke of the data within the data:

[0071]

[0072] in, Let i be the characteristic time series data of the i-th time interval, i∈(1,2,…,n), and n be the total number of time intervals;

[0073] When ke > 0, the data shows an upward trend, which may indicate that the equipment's operating condition is gradually deteriorating (such as an increase in temperature); when ke < 0, the data shows a downward trend; when ke ≈ 0, the data tends to stabilize, and the equipment is operating well.

[0074] For fluctuation characteristics, based on the characteristic time series data, the fluctuation characteristic coefficient Ca is calculated as follows: Ca = standard deviation of characteristic time series data / (absolute value of the mean of characteristic time series data + standard value of the monitoring indicator corresponding to the time series data). The standard value of the monitoring indicator corresponding to the time series data is not set to 0.

[0075] In this embodiment, the standard value of the monitoring indicator corresponding to the time series data is set according to the average value of the corresponding monitoring indicator when the equipment is running normally, and is not set to 0; when the average value of the monitoring indicator corresponding to the equipment is 0 when the equipment is running normally, the standard value of the monitoring indicator corresponding to the time series data is set to 10. -6 ;

[0076] Based on the characteristic time series data, the mutation characteristic coefficients for the k-th time interval are calculated to identify mutation features.

[0077] Where k∈(2,3,…,n-1), For the characteristic time series data of the (k-1)th time interval, This is the timing compensation constant. For the characteristic time series data of the (k+1)th time interval, This is a timing compensation constant, set to 10 in this embodiment. -6 .

[0078] In this embodiment, the mutation threshold is set according to the equipment type and historical data, and is generally set to 1.5-3. When it is greater than the threshold, it is considered that the data at that time point has undergone a mutation, which may indicate a potential equipment failure.

[0079] In one embodiment of the present invention, constructing an anomaly assessment model includes the following steps:

[0080] Obtain historical data and statistically analyze the trend characteristic coefficient, fluctuation characteristic coefficient, and sudden change characteristic coefficient of the equipment;

[0081] Based on the historical operating status of the equipment, the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that is operating normally and is not in the preset time interval before the failure are marked as normal operation; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that has failed are marked as failure; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment in the preset time interval before the failure are marked as potential failure.

[0082] LSTM models are trained for different monitoring indicators to identify the real-time operating status of electrical equipment.

[0083] In one embodiment of the present invention, fusing forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data includes the following steps:

[0084] Let the meteorological forecast data be X. m The meteorological monitoring data for the power grid is X. g Let w be the weight of the meteorological forecast data. m The weight of the power grid meteorological monitoring data is wg And w m +w g =1;

[0085] Weighted fusion of meteorological data Among them, the meteorological department's forecast data parameter w m =exp(-(t) m -t g ) / t0), t m t represents the time difference between the latest update time of meteorological data and the current time, as predicted by the meteorological department. g t0 represents the time difference between the latest update time of the meteorological data in the power grid meteorological monitoring data and the current time, and t0 represents the basic time interval for updating the meteorological data in the meteorological department's forecast data.

[0086] Statistical analysis of meteorological forecast data parameters w based on historical data m The maximum value of ′ for w m After normalization, the weights of the meteorological forecast data are obtained as w. m .

[0087] In one embodiment of the present invention, key meteorological influencing factors are selected for different equipment types, and a disaster impact quantification index is constructed based on fused meteorological data, including the following steps:

[0088] Based on fused meteorological data, key meteorological influencing factors were selected for different equipment types:

[0089] For transmission lines, the key meteorological influencing factors are set as wind speed, temperature, humidity, and lightning density;

[0090] For power generation equipment, the key meteorological influencing factors are set as temperature, precipitation, and lightning density;

[0091] For power equipment and distribution equipment, the key meteorological influencing factors are set as temperature, humidity and precipitation;

[0092] The formula for calculating the impact index of key meteorological influencing factors is set as follows:

[0093]

[0094] Among them, Ih is the impact index of key meteorological influencing factors, and th j Let tj be the duration of the j-th fused data for the key meteorological impact factors, th0 be the basic update time interval for the key meteorological impact factors, and xh be the duration of the data for the j-th fused data for the key meteorological impact factors. j Let xh be the value of the j-th fused data for the key meteorological influencing factors. min xh represents the minimum historical value of the key meteorological influencing factors.max is the maximum value of the historical data of the key meteorological impact factor, j∈(1,2,…,m), and m is the total number of updates of the fused data of the key meteorological impact factor;

[0095] The formula for the disaster impact quantification index is set as a weighted average formula for the impact index of key meteorological impact factors corresponding to different equipment types.

[0096] In this embodiment, for transmission lines, the key meteorological influencing factors are wind speed, temperature, humidity and lightning density, and the weights of the corresponding key meteorological influencing factor indices are set to 0.2, 0.3, 0.2 and 0.3 respectively.

[0097] For power generation equipment, the key meteorological influencing factors are temperature, precipitation and lightning density, and the weights of the corresponding key meteorological influencing factor indices are set to 0.3, 0.3 and 0.4 respectively.

[0098] For power equipment and power distribution equipment, the key meteorological influencing factors are temperature, humidity and precipitation, and the weights of the corresponding key meteorological influencing factor indices are set to 0.4, 0.3 and 0.3 respectively.

[0099] In one embodiment of the present invention, a multivariate time series is constructed by combining a disaster impact quantification index, operational anomaly assessment data, and power equipment operation data; the multivariate time series is then coarse-grained to obtain coarse-grained sequences at different time scales, including the following steps:

[0100] For each variable dimension of the multivariate time series consisting of disaster impact quantification index, operational anomaly assessment data, and power equipment operation data, a coarse-grained sequence is generated according to a preset time scale factor.

[0101] The coarse-grained sequence is obtained under the preset time scale factor s of the variable dimensions: Y(s)(g)={Y1(s)(g),Y2(s)(g),…,Y r (s)(g)};

[0102] Based on the coarse-grained sequence Y(s)(g) of the g-th variable dimension and the coarse-grained sequence Y(s)(h)={Y1(s)(h),Y2(s)(h),…,Y r By calculating the similarity matrix between the g-th and h-th variable dimensions (s)(h)}, we obtain the similarity matrix D among the multiple variables. s ;

[0103] Elements in the matrix , where Y a (s)(g) represents the a-th coarse-grained sequence parameter of the g-th variable dimension under the time scale factor s, Y a(s)(h) represents the a-th coarse-grained sequence parameter of the h-th variable dimension under the time scale factor s, where a∈(1,2,…,r), and r is the total number of coarse-grained sequence parameters of the g-th variable dimension under the time scale factor s.

[0104] In one embodiment of the present invention, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed based on coarse-grained sequences at different time scales to calculate the improved MMSE value of power equipment, including the following steps:

[0105] For the similarity matrix D between multivariates of the coarse-grained sequence at each time scale s, s Based on the adaptive similarity tolerance parameter rg, the matching probability C of dimension ds at time scale s is calculated. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r);

[0106] Among them, C s (ds)(r)= ; D is a similarity matrix of dimension ds among multiple variables. s In the middle The element, R s For similarity matrix D s The adaptive similarity tolerance parameter is calculated, where Θ() is the Heavgsgdse step function and N0 is the sequence length; similarly, C is calculated. s (ds+1)(r);

[0107] An improved multivariate multiscale sample entropy (MMSE) calculation model is constructed, based on the matching probability C with dimension ds. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r) is used to calculate the multivariate multiscale sample entropy, which serves as the improved MMSE value for power equipment: MMSE(r,s,R) s ,N0)=-ln(C s (ds+1)(r) / C s (ds)(r));

[0108] Where r is the total number of coarse-grained sequence parameters in the g-th variable dimension under the time scale factor s.

[0109] In one embodiment of the present invention, a fault feature vector is constructed based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales; by analyzing the distribution characteristics of the fault feature vectors under different fault types, power equipment fault classification is performed, including the following steps:

[0110] The improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales are combined with trend characteristic coefficients, fluctuation characteristic coefficients and abrupt change characteristic coefficients to form a high-dimensional fault feature vector;

[0111] The ReliefF algorithm is used to remove redundant features, and the distribution characteristics of high-dimensional fault feature vectors under different fault types are analyzed by box plots or t-SNE dimensionality reduction techniques.

[0112] For example, the feature vector of a short-circuit fault may exhibit a distribution with high kurtosis and low permutation entropy, while an overload fault will exhibit characteristics of high mean and high sample entropy.

[0113] The LSTM (Long Short-Term Memory) network model is trained by analyzing the distribution characteristics of high-dimensional fault feature vectors under different fault types in historical data. The spatiotemporal correlation of the distribution characteristics is learned, and the trained LSTM network model is used to classify power equipment faults.

[0114] In this embodiment, the distribution characteristics of high-dimensional fault feature vectors under different fault types are analyzed by using box plots or t-SNE dimensionality reduction techniques.

[0115] To address the high-dimensional fault feature vectors of power equipment, MEMD multidimensional empirical mode decomposition is used to adaptively decompose non-stationary signals. MEMD obtains the IMF multivariate intrinsic mode functions of the signal through multi-directional projection.

[0116] By utilizing the multivariate IMF components of MEMD decomposition, we can construct cross-variable entropy value combinations for different fault types to enhance the distinguishability of fault types.

[0117] In another embodiment of the present invention, a power equipment data intelligent analysis system based on 5G communication is provided, comprising:

[0118] Data Acquisition Layer:

[0119] Based on the type of power equipment, the power equipment is divided into several subcategories. Operational data of the power equipment in each subcategory is collected, and geographical location data and local meteorological data of the power equipment in each subcategory are obtained.

[0120] Communication transport layer:

[0121] Based on the distribution of power equipment, the entire power area is divided into multiple sub-areas. Each sub-area is equipped with a 5G communication base station to transmit the data collected by the data acquisition layer to a data relay node. The data relay node regroups the acquired data according to the type of power equipment and sends the grouped data to the cloud analysis layer. In this embodiment, the location data of the power equipment is reverse-located using the 5G communication base station and compared with the geographical location data of the power equipment collected by the data acquisition layer. Data in the original location data that deviates from the reverse-located power equipment location data by more than a threshold is replaced.

[0122] The operation data of each group of power equipment is standardized. Based on the time series data of each standardized monitoring indicator, the characteristic coefficient is analyzed according to the characteristic indicator formula to construct an operation anomaly assessment model and send early warning signals to the power equipment with operation anomalies in the corresponding sub-region.

[0123] Cloud analytics layer:

[0124] The forecast data provided by the meteorological department is fused with the power grid meteorological monitoring data collected by meteorological monitoring sensors around the power equipment to obtain fused meteorological data; for different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on the fused meteorological data;

[0125] The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales.

[0126] Based on coarse-grained sequences at different time scales, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed to calculate the improved MMSE value of power equipment.

[0127] Based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales, fault feature vectors are constructed; by analyzing the distribution characteristics of fault feature vectors under different fault types, power equipment faults are classified.

[0128] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent analysis of power equipment data based on 5G communication, characterized in that, Includes the following steps: Based on the type of power equipment, the power equipment is divided into several subcategories. Operational data of the power equipment in each subcategory is collected, and geographical location data and local meteorological data of the power equipment in each subcategory are obtained. Based on the distribution of power equipment, the entire power area is divided into multiple sub-areas, and each sub-area is equipped with a 5G communication base station to transmit operational data and meteorological data to the data relay node; The data relay node will regroup the acquired data according to the type of power equipment and send the grouped data to the cloud; The data transfer node analyzes the characteristic coefficients according to the characteristic index formula based on the time series data of different monitoring indicators of the operation data, constructs the operation anomaly assessment model, and sends early warning signals to the power equipment with operation anomalies in the corresponding sub-region. The cloud-based analysis system integrates forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data. For different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on fused meteorological data; The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales. An improved multivariate multiscale sample entropy MMSE calculation model was constructed to calculate the improved MMSE value of power equipment. Based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales, a fault feature vector is constructed. Power equipment faults are classified by analyzing the distribution characteristics of fault feature vectors under different fault types. The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series. The multivariate time series is then coarse-grained to obtain coarse-grained sequences at different time scales, including the following steps: For each variable dimension of the multivariate time series consisting of disaster impact quantification index, operational anomaly assessment data, and power equipment operation data, a coarse-grained sequence is generated according to a preset time scale factor. The coarse-grained sequence is obtained under the preset time scale factor s of the variable dimensions: Y(s)(g)={Y1(s)(g),Y2(s)(g),…,Y r (s)(g)}; Based on the coarse-grained sequence Y(s)(g) of the g-th variable dimension and the coarse-grained sequence Y(s)(h)={Y1(s)(h),Y2(s)(h),…,Y r By calculating the similarity matrix between the g-th and h-th variable dimensions (s)(h)}, we obtain the similarity matrix D among the multiple variables. s ; Elements in the matrix , where Y a (s)(g) represents the a-th coarse-grained sequence parameter of the g-th variable dimension under the time scale factor s, Y a (s)(h) represents the a-th coarse-grained sequence parameter of the h-th variable dimension under the time scale factor s, where a∈(1,2,…,r), and r is the total number of coarse-grained sequence parameters of the g-th variable dimension under the time scale factor s; An improved multivariate multiscale sample entropy MMSE calculation model is constructed to calculate the improved MMSE value of power equipment, including the following steps: For the similarity matrix D between multivariates of the coarse-grained sequence at each time scale s, s Based on the adaptive similarity tolerance parameter rg, the matching probability C of dimension ds at time scale s is calculated. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r); Among them, C s (ds)(r)= ; D is a similarity matrix of dimension ds among multiple variables. s In the middle The element, R s For similarity matrix D s The adaptive similarity tolerance parameter, Let the Heavgsgdse step function be the sequence length, and N0 be the sequence length; similarly, calculate C. s (ds+1)(r); An improved multivariate multiscale sample entropy (MMSE) calculation model is constructed, based on the matching probability C with dimension ds. s (ds)(r), and C when the dimension is ds+1 s (ds+1)(r) is used to calculate the multivariate multiscale sample entropy, which serves as the improved MMSE value for power equipment: MMSE(r,s,R) s ,N0)=-ln(C s (ds+1)(r) / C s (ds)(r)); Where r is the total number of coarse-grained sequence parameters in the g-th variable dimension under the time scale factor s; Power equipment fault classification is performed by analyzing the distribution characteristics of fault feature vectors under different fault types, including the following steps: The improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales are combined with trend characteristic coefficients, fluctuation characteristic coefficients and abrupt change characteristic coefficients to form a high-dimensional fault feature vector; The ReliefF algorithm is used to remove redundant features, and the distribution characteristics of high-dimensional fault feature vectors under different fault types are analyzed by box plots or t-SNE dimensionality reduction techniques. The LSTM (Long Short-Term Memory) network model is trained by analyzing the distribution characteristics of high-dimensional fault feature vectors under different fault types in historical data. The spatiotemporal correlation of the distribution characteristics is learned, and the trained LSTM network model is used to classify power equipment faults.

2. The intelligent analysis method for power equipment data based on 5G communication according to claim 1, characterized in that, The subcategories include: transmission lines, power generation equipment, transformer equipment, and distribution equipment; the operating data of power equipment includes: current, voltage, power, temperature, humidity, and amplitude; the meteorological data types include: temperature, humidity, wind speed, precipitation, and lightning density. The meteorological data includes meteorological data collected by meteorological monitoring sensors around the power equipment, as well as forecast data provided by the meteorological department.

3. The intelligent analysis method for power equipment data based on 5G communication according to claim 1, characterized in that, Based on time-series data of different monitoring indicators in operational data, characteristic coefficient analysis is performed according to the characteristic indicator formula, including the following steps: The operating data of each group of power equipment is standardized, and the trend characteristic coefficient, fluctuation characteristic coefficient and sudden change characteristic coefficient are analyzed from the time series data of each monitoring indicator of the standardized power equipment operating data. For any monitoring indicator, set the time period [t] for the operational data monitoring indicator. a ,t b The time interval is divided into different intervals, with corresponding time interval numbers (1, 2, ..., n). The standardized mean of the data within each time interval is taken as the feature time series data: (x1′, x2′, ..., x n ′); Based on the trend characteristics and the characteristic time series data, the time interval [t] is calculated. a ,t b The linear trend slope characteristic coefficient ke of the data within the data: ； in, Let i be the characteristic time series data of the i-th time interval, i∈(1,2,…,n), and n be the total number of time intervals; For fluctuation characteristics, based on the characteristic time series data, the fluctuation characteristic coefficient Ca is calculated as follows: Ca = standard deviation of characteristic time series data / (absolute value of the mean of characteristic time series data + standard value of the monitoring indicator corresponding to the time series data). The standard value of the monitoring indicator corresponding to the time series data is not set to 0. Based on the characteristic time series data, the mutation characteristic coefficients for the k-th time interval are calculated to identify mutation features. ； Where k∈(2,3,…,n-1), For the characteristic time series data of the (k-1)th time interval, This is the timing compensation constant.

4. The intelligent analysis method for power equipment data based on 5G communication according to claim 3, characterized in that, Constructing an anomaly assessment model includes the following steps: Obtain historical data and statistically analyze the trend characteristic coefficient, fluctuation characteristic coefficient, and sudden change characteristic coefficient of the equipment; Based on the historical operating status of the equipment, the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that is operating normally and is not in the preset time interval before the failure are marked as normal operation; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment that has failed are marked as failure; the trend characteristic coefficients, fluctuation characteristic coefficients, and sudden change characteristic coefficients of the equipment in the preset time interval before the failure are marked as potential failure. LSTM models are trained for different monitoring indicators to identify the real-time operating status of electrical equipment.

5. The intelligent analysis method for power equipment data based on 5G communication according to claim 1, characterized in that, The process of fusing forecast data provided by meteorological departments with power grid meteorological monitoring data collected by meteorological monitoring sensors around power equipment to obtain fused meteorological data includes the following steps: Let the meteorological forecast data be X. m The meteorological monitoring data for the power grid is X. g Let w be the weight of the meteorological forecast data. m The weight of the power grid meteorological monitoring data is w g And w m +w g =1; Weighted fusion of meteorological data Among them, the meteorological department's forecast data parameter w m =exp(-(t) m -t g ) / t0), t m t represents the time difference between the latest update time of meteorological data and the current time, as predicted by the meteorological department. g t0 represents the time difference between the latest update time of the meteorological data in the power grid meteorological monitoring data and the current time, and t0 represents the basic time interval for updating the meteorological data in the meteorological department's forecast data. Statistical analysis of meteorological forecast data parameters w based on historical data m The maximum value of ′ for w m After normalization, the weights of the meteorological forecast data are obtained as w. m .

6. The intelligent analysis method for power equipment data based on 5G communication according to claim 1, characterized in that, For different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on fused meteorological data, including the following steps: Based on fused meteorological data, key meteorological influencing factors were selected for different equipment types: For transmission lines, the key meteorological influencing factors are set as wind speed, temperature, humidity, and lightning density; For power generation equipment, the key meteorological influencing factors are set as temperature, precipitation, and lightning density; For power equipment and distribution equipment, the key meteorological influencing factors are set as temperature, humidity and precipitation; The formula for calculating the impact index of key meteorological influencing factors is set as follows: ； Among them, Ih is the impact index of key meteorological influencing factors, and th j Let tj be the duration of the j-th fused data for the key meteorological impact factors, th0 be the basic update time interval for the key meteorological impact factors, and xh be the duration of the data for the j-th fused data for the key meteorological impact factors. j Let xh be the value of the j-th fused data for the key meteorological influencing factors. min xh represents the minimum historical value of the key meteorological influencing factors. max is the maximum value of the historical data of the key meteorological impact factor, j∈(1,2,…,m), and m is the total number of updates of the fused data of the key meteorological impact factor; The formula for the disaster impact quantification index is set as a weighted average formula for the impact index of key meteorological impact factors corresponding to different equipment types.

7. A power equipment data intelligent analysis system based on 5G communication, characterized in that, This system employs a 5G-based intelligent analysis method for power equipment data, as described in any one of claims 1-6, to achieve intelligent analysis of power equipment data, including: Data Acquisition Layer: Based on the type of power equipment, the power equipment is divided into several subcategories. Operational data of the power equipment in each subcategory is collected, and geographical location data and local meteorological data of the power equipment in each subcategory are obtained. Communication transport layer: Based on the distribution of power equipment, the entire power area is divided into multiple sub-areas. Each sub-area is equipped with a 5G communication base station to transmit the data collected by the data acquisition layer to the data relay node. The data relay node then groups the acquired data according to the type of power equipment and sends the grouped data to the cloud analysis layer. The operation data of each group of power equipment is standardized. Based on the time series data of each standardized monitoring indicator, the characteristic coefficient is analyzed according to the characteristic indicator formula to construct an operation anomaly assessment model and send early warning signals to the power equipment with operation anomalies in the corresponding sub-region. Cloud analytics layer: The forecast data provided by the meteorological department is fused with the power grid meteorological monitoring data collected by meteorological monitoring sensors around the power equipment to obtain fused meteorological data; for different equipment types, key meteorological influencing factors are selected, and a disaster impact quantification index is constructed based on the fused meteorological data; The disaster impact quantification index, operational anomaly assessment data, and power equipment operation data are used to construct a multivariate time series; the multivariate time series is then coarse-grained to obtain coarse-grained series at different time scales. Based on coarse-grained sequences at different time scales, an improved multivariate multiscale sample entropy (MMSE) calculation model is constructed to calculate the improved MMSE value of power equipment. Based on the improved MMSE values ​​of multivariate monitoring time series of power equipment at different time scales, fault feature vectors are constructed; by analyzing the distribution characteristics of fault feature vectors under different fault types, power equipment faults are classified.

Images