Electric energy meter health assessment method and system based on operation state

Abstract

The invention discloses an electric energy meter health assessment method and system based on an operation state, and relates to the technical field of electric energy metering, and the method comprises the steps: collecting the electrical operation parameters and temperature monitoring parameters of an electric energy meter in real time; extracting chip real-time temperature characteristic data based on chip temperature related data, constructing an electric energy meter operation state space in combination with the electrical operation parameters, and establishing a state transition behavior model; performing feature learning on the state transition behavior model through historical normal operation data, and extracting standard state transition features; inputting the collected real-time operation state parameters into a state transition behavior model to obtain actual state transition characteristics; and comparing the difference degree between the actual state transition feature and the standard state transition feature, and generating electric energy meter health state prompt information. According to the invention, the technical problem that the hidden fault of the electric energy meter is difficult to accurately and early recognize in the prior art is solved, and the technical effects of accurate evaluation of the health state of the electric energy meter and early warning of the hidden fault are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric energy metering, in particular to an electric energy meter health assessment method and system based on operating state. BACKGROUND

[0002] During long-term operation, electric energy meters are susceptible to influences of various factors such as voltage fluctuation, load change and temperature aging, and internal components may gradually produce performance degradation or potential abnormalities, but such implicit faults usually do not manifest as obvious metering errors or functional failures in the early stage. Traditional health assessment methods mostly rely on single operating parameter monitoring or fixed threshold judgment, lack comprehensive analysis capability for multi-dimensional operating state and its dynamic evolution process, and are difficult to identify subtle abnormal changes in state transition process, thereby making it difficult to discover implicit faults in a timely and accurate manner. SUMMARY

[0003] The present application provides an electric energy meter health assessment method and system based on operating state, which is used to solve the technical problem that it is difficult to accurately and early identify implicit faults of electric energy meters in the prior art.

[0004] In view of the above problems, the present application provides an electric energy meter health assessment method and system based on operating state.

[0005] In a first aspect of the present application, an electric energy meter health assessment method based on operating state is provided, which comprises:

[0006] Real-time acquisition of electrical operating parameters and temperature monitoring parameters of the electric energy meter, wherein the temperature monitoring parameters include chip temperature related data and environmental temperature data; extraction of chip real-time temperature feature data based on the chip temperature related data, combination of the electrical operating parameters, construction of an electric energy meter operating state space, and establishment of a state transition behavior model based on the electric energy meter operating state space; feature learning of the state transition behavior model through historical normal operating data, extraction of standard state transition features; input of the acquired real-time operating state parameters into the state transition behavior model to obtain actual state transition features; comparison of the difference degree between the actual state transition features and the standard state transition features, and generation of electric energy meter health state prompt information when the difference degree exceeds a preset difference threshold.

[0007] In a second aspect of the present application, an electric energy meter health assessment system based on operating state is provided, which comprises: The data acquisition module is used for collecting electrical operation parameters and temperature monitoring parameters of the electric energy meter in real time, the temperature monitoring parameters include chip temperature related data and environmental temperature data; the model establishment module is used for extracting chip real-time temperature feature data based on the chip temperature related data, combining the electrical operation parameters, constructing an electric energy meter operation state space, and establishing a state transition behavior model based on the electric energy meter operation state space; the feature learning module is used for performing feature learning on the state transition behavior model through historical normal operation data, and extracting standard state transition features; the transition feature acquisition module is used for inputting collected real-time operation state parameters into the state transition behavior model to obtain actual state transition features; and the information generation module is used for comparing the difference degree of the actual state transition features and the standard state transition features, and generating an electric energy meter health state prompt information when the difference degree exceeds a preset difference threshold.

[0008] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: The present application collects electrical operation parameters and temperature monitoring parameters of the electric energy meter in real time, the temperature monitoring parameters include chip temperature related data and environmental temperature data; extracts chip real-time temperature feature data based on the chip temperature related data, combines the electrical operation parameters, constructs an electric energy meter operation state space, and establishes a state transition behavior model based on the electric energy meter operation state space; performs feature learning on the state transition behavior model through historical normal operation data, and extracts standard state transition features; inputs collected real-time operation state parameters into the state transition behavior model to obtain actual state transition features; and compares the difference degree of the actual state transition features and the standard state transition features, and generates an electric energy meter health state prompt information when the difference degree exceeds a preset difference threshold. The present application solves the technical problem that it is difficult to accurately and early identify the implicit failure of the electric energy meter in the prior art, and achieves the technical effects of accurate evaluation of the health state of the electric energy meter and early warning of the implicit failure by constructing the operation state space and the state transition behavior model and comparing the standard and actual state transition features. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0010] Figure 1 A flowchart of an electric energy meter health evaluation method based on an operation state provided by the embodiments of the present application is shown in the figure. Figure 2A structure schematic diagram of a health assessment system of an electric energy meter based on an operating state is provided in the embodiments of the present application.

[0011] Marker explanation: data acquisition module 11, model establishment module 12, feature learning module 13, transition feature acquisition module 14, information generation module 15. DETAILED DESCRIPTION

[0012] The present application provides a health assessment method and system of an electric energy meter based on an operating state, aiming to solve the technical problem that it is difficult to accurately and early identify the hidden faults of the electric energy meter in the prior art. By constructing an operating state space and a state transition behavior model and comparing the standard and actual state transition features, the technical effect of accurately assessing the health state of the electric energy meter and early warning the hidden faults is achieved.

[0013] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0014] It should be noted that any variation of the terms "comprise" and "have" is intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to those clearly listed steps or units, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0015] Embodiment one, as shown in the present application provides a health assessment method of an electric energy meter based on an operating state, which comprises: Figure 1 Step S100: Real-time acquisition of electrical operating parameters and temperature monitoring parameters of the electric energy meter, wherein the temperature monitoring parameters comprise chip temperature related data and environmental temperature data. In the embodiments of the present application, during the operation of the electric energy meter, the metering circuit of the electric energy meter is first monitored online. The voltage signal is acquired through the voltage sampling channel, the current signal is acquired through the current sampling channel corresponding to the current transformer or shunt, and the sampling signal is filtered, amplified and analog-digital converted, so as to form the electrical operating parameters corresponding to the real-time working state of the electric energy meter. The electrical operating parameters are used to represent the voltage, current and their operating characteristics changing with time, and reflect the load change and the operation of the metering circuit.

[0016]

[0017] ​Meanwhile, the temperature state of the electric energy meter is monitored online, a temperature sensor is arranged near the metering chip, temperature measurement values corresponding to the thermal state changes of the metering chip are read in real time, and chip temperature related data are obtained; the chip temperature related data are used to represent the temperature rise and thermal accumulation characteristics generated by the metering chip during operation. A temperature detection element is arranged inside the shell of the electric energy meter or at a mounting environment position, the surrounding air temperature or the shell surface temperature is read in real time, and environment temperature data are obtained; the environment temperature data are used to represent the external thermal conditions and environment temperature fluctuation of the electric energy meter.

[0018] Then, the electrical operation parameters, the chip temperature related data and the environment temperature data are time-synchronized and data-aligned, so that each parameter corresponds to each other at the same sampling time, so as to obtain the electrical operation parameters and the temperature monitoring parameters.

[0019] Step S200: extracting chip real-time temperature feature data based on the chip temperature related data, combining the electrical operation parameters, constructing an electric energy meter operation state space, and establishing a state transition behavior model based on the electric energy meter operation state space.

[0020] In the embodiments of the present application, when the chip real-time temperature feature data is extracted based on the chip temperature related data, first, the chip temperature related data are subjected to temperature sensitive feature extraction processing to obtain primary temperature feature data; second, the primary temperature feature data are subjected to temperature state calculation processing in combination with the circuit board thermal conduction characteristic data to obtain intermediate temperature feature data; and finally, the intermediate temperature feature data are subjected to aging feature extraction processing according to a temperature accumulation effect analysis model to generate the chip real-time temperature feature data.

[0021] Next, the electric energy meter operation state space is constructed in combination with the electrical operation parameters. In this process, first, the electrical operation parameters and the chip real-time temperature feature data are subjected to parameter fusion processing to obtain state space basic data; second, state change modes are defined based on the state space basic data to form state space change description data; and finally, the state space basic data and the state space change description data are subjected to state transition rule analysis in combination with device historical operation data to construct the electric energy meter operation state space.

[0022] Finally, the state transition behavior model is established based on the electric energy meter operation state space. In this process, first, the state space basic data and the state space change description data are obtained from the electric energy meter operation state space; second, a state transition probability description matrix containing state transition conditions and probability values is constructed based on the state space basic data and the state space change description data; and finally, the state transition behavior model is established according to the state transition probability description matrix, so as to realize probabilistic modeling of the electric energy meter operation state evolution process.

[0023] Further, the method provided by the application embodiment further comprises the following steps: The temperature sensitive feature extraction processing is performed on the chip temperature related data to obtain primary temperature feature data; the temperature state calculation processing is performed on the primary temperature feature data based on the circuit board heat conduction characteristic data to obtain intermediate temperature feature data; and the aging feature extraction processing is performed on the intermediate temperature feature data according to a temperature cumulative effect analysis model to obtain chip real-time temperature feature data.

[0024] In the application embodiment, when the temperature sensitive feature extraction processing is performed on the chip temperature related data, first, the chip temperature related data is arranged in the order of sampling time, and the missing points are linearly interpolated to complete the missing points. The temperature difference between the two known sampling points is distributed to the missing time according to the time proportion, so that the missing point temperature falls on the position corresponding to the temperature connecting line of the two ends. Then, the abnormal jump correction is performed. The temperature difference between each sampling point and the previous sampling point is calculated. When the absolute value of the difference value exceeds the preset jump threshold, the temperature of the sampling point is corrected to the average value of the temperatures of the adjacent sampling points before and after it, so that the temperature sequence is continuous and has no mutation. After the completion of the interpolation and correction, the sliding time window is segmented. The continuous temperature sequence is divided into multiple windows according to the fixed window length. In each window, the temperatures of all sampling points in the window are added one by one and divided by the number of sampling points to obtain the window average temperature. At the same time, the difference between the average temperatures of the adjacent two windows is calculated as the window change amount. Finally, the window average temperature and the window change amount of each window are output as the primary temperature feature data.

[0025] When the temperature state calculation processing is performed on the primary temperature feature data based on the circuit board heat conduction characteristic data, first, the heat conductivity parameter, the thermal resistance parameter and the heat capacity parameter pre-stored in the circuit board heat conduction characteristic data are read, and the heat conduction path is represented as a thermal resistance network according to the circuit board structure, and each section of material or structure in the heat conduction path is corresponded to a thermal resistance element. Then, the equivalent thermal resistance is calculated according to the connection relationship of the thermal resistance network. If it is a series path, the thermal resistances of each section are added to obtain the total thermal resistance. If it is a parallel path, the reciprocals of the branch thermal resistances are summed and then the reciprocal is taken to obtain the equivalent thermal resistance. Then, the temperature difference between the window average temperature and the environment temperature is taken as the total temperature difference. In the series path, the temperature difference of each section is equal to the total temperature difference multiplied by the ratio of the thermal resistance of the section to the total thermal resistance. In the parallel path, the temperature differences at both ends of each branch are the same and are determined by the equivalent thermal resistance relationship. After obtaining the temperature difference of each path, the temperature state value of the corresponding path position is obtained by adding the environment temperature and the path temperature difference, thereby forming the intermediate temperature feature data.

[0026] According to the temperature cumulative effect analysis model, when the intermediate temperature characteristic data is subjected to aging characteristic extraction processing, first, the temperature cumulative effect analysis model is established and the model parameters are determined. In the establishment process, the sample data is formed by collecting the accelerated life test data and operation record data of the same type of electric energy meter under different temperature levels and different duration conditions, the sample data including the intermediate temperature characteristic data, the duration information corresponding to the intermediate temperature characteristic data, and the measurement values of the electrical performance parameters at the test start time and the test end time, the measurement values of the electrical performance parameters including at least the power supply current value of the metering chip, the internal reference voltage value, and the measurement error value. Then, the aging degree value is calculated based on the measurement values of the electrical performance parameters, specifically, the ratio of the power supply current change amount to the initial value of the power supply current, the ratio of the internal reference voltage change amount to the initial value of the internal reference voltage, and the absolute value of the measurement error change amount are calculated, and then the above three items are added according to the preset weight coefficient to obtain the sample aging degree value. At the same time, the sample temperature cumulative amount is calculated based on the intermediate temperature characteristic data and the duration information in the sample, the sample is divided into a plurality of continuous time periods, the temperature state value and the duration of each time period are recorded, the temperature state value and the duration of each time period are multiplied to obtain the temperature cumulative amount of the time period, and the temperature cumulative amounts of the time periods are added sequentially to obtain the sample temperature cumulative amount. Then, the model parameters of the temperature cumulative effect analysis model are determined by curve fitting and least squares method, so that the error between the model output calculated from the sample temperature cumulative amount and the sample aging degree value satisfies the preset error condition, thereby obtaining the temperature cumulative effect analysis model with determined parameters, and the temperature cumulative effect analysis model satisfies the calculation relationship that the aging degree value is equal to the power mapping of the model parameters and the temperature cumulative amount and the superposition of the bias term.

[0027] After the parameters are determined, the intermediate temperature characteristic data obtained in real time is divided into a plurality of continuous time periods in the order of sampling time, the temperature state value and the duration of each time period are recorded, the real-time temperature cumulative amount is calculated, the temperature state value and the duration of each time period are multiplied to obtain the temperature cumulative amount of the time period, and the temperature cumulative amounts of the time periods are added sequentially to obtain the real-time temperature cumulative amount; then the real-time temperature cumulative amount is input into the temperature cumulative effect analysis model with determined parameters, the real-time temperature cumulative amount is subjected to power mapping and superposition of the bias term according to the model parameters, and the real-time aging degree value is obtained; finally, the real-time aging degree value is output as the real-time temperature characteristic data of the chip.

[0028] Further, the method provided by the application embodiment further comprises: perform parameter fusion processing on the electrical operation parameters and the chip real-time temperature feature data to obtain state space basic data; defining a state change mode based on the state space basic data to obtain state space change description data; performing state transition rule analysis on the state space basic data and the state space change description data according to historical operation data of the device to construct an operation state space of the electric energy meter.

[0029] In the embodiments of the present application, when the electrical operation parameters and the chip real-time temperature feature data are subjected to parameter fusion processing, firstly, time synchronization and sampling point matching are performed on the electrical operation parameters and the chip real-time temperature feature data, so that the electrical operation parameters such as voltage, current, power factor and frequency at the same sampling time are one-to-one corresponding to the chip real-time temperature feature data at the same sampling time; then, dimension unification processing is performed on each parameter, and the minimum-maximum normalization method is adopted to linearly scale each parameter according to the minimum value and the maximum value determined by the historical normal operation data, to obtain normalized electrical operation parameters and normalized chip real-time temperature feature data; after the normalization processing is completed, the normalized electrical operation parameters and the normalized chip real-time temperature feature data are combined into a state feature vector of the same dimension according to a preset parameter arrangement order by using a feature vector splicing method, and the state feature vector is output as the state space basic data, which is used to represent the operation state value of the electric energy meter at a single time.

[0030] When the state change mode is defined based on the state space basic data, firstly, the state feature vectors at consecutive sampling times are extracted to form a time sequence, and the state feature vectors at adjacent sampling times are subjected to difference calculation to obtain a state increment vector; then, interval binning processing is performed on the state space basic data, that is, all sample values of each parameter are counted in the historical normal operation data, sorted in ascending order of value, and the corresponding percentile positions are calculated, for example, the 25th percentile, the 50th percentile and the 75th percentile corresponding parameter values are calculated, and the percentile corresponding values are taken as the bin boundaries, so that the continuous values are divided into multiple intervals; for example, the voltage parameter has a value range of 180 volts to 260 volts in the historical normal operation data, and after sorting, the 25th percentile is 200 volts, the 50th percentile is 220 volts, and the 75th percentile is 240 volts, so the voltage parameter is divided into four intervals of less than 200 volts, 200 volts to 220 volts, 220 volts to 240 volts and greater than 240 volts; after each parameter is divided into intervals according to the above method, the interval numbers of each parameter are combined according to a preset encoding rule to form a discrete state code; after obtaining the discrete state code, the discrete state code at the current sampling time is corresponded to the discrete state code at the next sampling time, the change is defined as a state transition, and the state increment vector and the transition direction corresponding to the state transition are recorded to form the state space change description data.

[0031] According to the device historical operation data, when the state transition rule analysis is performed on the state space basic data and the state space change description data, firstly, the time sequence of the discrete state code is extracted from the historical operation data, and the number of transitions from the current discrete state code to the next discrete state code is obtained by counting each pair of adjacent discrete state codes; then, the total transition number of the discrete state code is obtained by summing up all the transition numbers of the same current discrete state code, and the corresponding state transition probability is obtained by dividing each transition number by the total transition number; the obtained state transition probability is arranged according to the current discrete state code as the row and the next discrete state code as the column to form a state transition probability matrix; finally, the discrete state code system and the state transition probability matrix are collectively represented as the electric energy meter operation state space.

[0032] Further, the method provided by the application embodiment further comprises the following steps: obtaining state space basic data and state space change description data from the electric energy meter operation state space; constructing a state transition probability description matrix containing state transition conditions and probability values based on the state space basic data and the state space change description data; and establishing a state transition behavior model according to the state transition probability description matrix.

[0033] In the application embodiment, when the state space basic data and the state space change description data are obtained from the electric energy meter operation state space, firstly, the storage content of the constructed electric energy meter operation state space is read, and the state space basic data and the state space change description data are derived in chronological order; then, the two types of data are matched piece by piece with the sampling time as the index, the corresponding current discrete state code is determined for each sampling time, and the discrete state code of the next sampling time is read, the current discrete state code and the next discrete state code are combined to form a state transition record, and the state space basic data and the state space change description data corresponding to the state transition record are retained.

[0034] When the state transition probability description matrix containing the state transition conditions and the probability values is constructed based on the state space basic data and the state space change description data, firstly, all the state transition records are traversed, for each state transition record, the current discrete state code is taken as the row index, the next discrete state code is taken as the column index, and the count value is added one in the corresponding matrix position, so as to complete the accumulation of the state transition number; then, the total number of each row is obtained by summing up all the count values of each row, and the count value of each column of the row is divided by the total number of the row to obtain the probability value of the corresponding position, so as to convert the count matrix into the state transition probability description matrix, and the sum of the probability values of each row is one.

[0035] When the state transition probability description matrix is used to establish a state transition behavior model, the state transition probability description matrix is first written into the model as transition parameters, and a cumulative probability sequence is calculated for each row. The cumulative probability sequence is obtained by sequentially adding the probability values in each column of the row. Then, when performing state transition, the cumulative probability sequence corresponding to the current discrete state code is read, a random number between zero and one is generated, and the column index corresponding to the interval in which the random number falls is determined as the next discrete state code. The above steps are repeated with the next discrete state code as the new current discrete state code, so as to generate a continuous state transition sequence according to probability and form a state transition behavior model.

[0036] Further, the method provided by the application embodiment further comprises: Collecting historical running state transition sequence data, and based on the historical running state transition sequence data, counting the transition frequency and transition condition between states, and constructing a state transition probability description matrix according to the transition frequency and transition condition, and setting a probability updating rule.

[0037] In the application embodiment, when collecting historical running state transition sequence data, the discrete state codes recorded by time stamp are first read from the electric energy meter running data storage medium, and the discrete state codes are resampled and aligned according to a unified sampling period. When there is a lack of discrete state codes at a certain sampling time, the discrete state codes at the previous sampling time are used to fill in to ensure the continuity of the sequence. Then, the discrete state codes are sorted in ascending order according to the time stamp and grouped according to the device number for output, so that each electric energy meter forms an independent discrete state code time sequence. After obtaining the discrete state code time sequence, the discrete state code at time t and the discrete state code at time t+1 are paired to form a state transition record, and the continuous state transition record sequence is obtained by sequentially traversing all adjacent time points, so as to obtain the historical running state transition sequence data. The historical running state transition sequence data is used as the original statistical data source of the state transition probability description matrix.

[0038] When the transition frequency and transition condition between states are counted based on the historical running state transition sequence data, first, a transition frequency counting table is established, taking the current discrete state code as the row index and the next discrete state code as the column index, and initializing each cell count to zero. Then each state transition record in the historical running state transition sequence data is traversed, the current discrete state code of the record is determined as the state transition condition, and the next discrete state code of the record is determined as the transition result. The corresponding row and column position in the transition frequency counting table is counted by one, thereby obtaining the transition frequency between states. After the traversal is completed, the current discrete state code corresponding to each row is determined as the state transition condition, and the column index and its count value in the row with a non-zero count value are taken as the effective transition result and its corresponding frequency under the state transition condition, thereby completing the counting of the transition condition and the transition frequency. The transition frequency counting result is used to construct the state transition probability description matrix.

[0039] When the state transition probability description matrix is constructed according to the transition frequency and the transition condition and the probability update rule is set, first, the transition frequency counting table is taken as the transition frequency matrix, and each row of the transition frequency matrix is normalized, that is, the total transition times of a current discrete state code are obtained by summing all the count values in the row corresponding to the current discrete state code. Then, the count value of each column in the row is divided by the total transition times to obtain the probability value of transitioning to the corresponding next discrete state code under the state transition condition, and the calculated probability value is used to replace the original count value, thereby forming the state transition probability description matrix, so that the sum of the probability values in each row of the matrix is equal to one. When a current discrete state code does not transition within the statistical period, to avoid a zero denominator, the probability value of the row is kept as zero or assigned according to a preset initial probability. The state transition probability description matrix is taken as the transition parameter input of the subsequent state transition behavior model.

[0040] In setting the probability updating rule, first, the incremental updating method is used to process the newly added historical running state transition record. When the newly added state transition record arrives, the corresponding current discrete state code and next discrete state code are determined according to the record, and the count value in the corresponding row and column position in the transition frequency matrix is added by one. Then, the total transition times of the row where the current discrete state code is located are recalculated, and the count values of each column in the row are divided by the new total transition times to update the probability values corresponding to the row. At the same time, the time window control strategy is used to limit the statistical range to the state transition records in the recent preset time length. When the earliest state transition record entering the statistical range exceeds the preset time length, the count value in the corresponding row and column position in the transition frequency matrix is reduced by one, and the total transition times and probability values of the row are recalculated, so as to ensure that the state transition probability description matrix can dynamically reflect the latest transition rule of the running state of the electric energy meter, so as to ensure that the data source and updating logic of the state transition probability description matrix are consistent in the model construction and running stages.

[0041] Step S300: learning features of the state transition behavior model through historical normal running data, and extracting standard state transition features.

[0042] In the embodiments of the present application, when learning features of the state transition behavior model through historical normal running data and extracting standard state transition features, first, the historical running data of the electric energy meter in the normal running state in a preset time period is selected to form feature learning sample data. Then, the feature learning sample data is input into the state transition behavior model to perform behavior feature reverse learning processing on the state transition path and transition probability distribution in the normal running state, so as to extract standard state transition features capable of representing the evolution rule of the normal running state.

[0043] Further, in the method provided by the embodiments of the present application, learning features of the state transition behavior model through historical normal running data and extracting standard state transition features further include: The historical normal running data in a preset time period is selected as feature learning sample data. The feature learning sample data is input into the state transition behavior model to perform behavior feature reverse learning processing, and standard state transition features are extracted.

[0044] In the embodiment of the present application, when the historical normal operation data in the preset time period is selected as the feature learning sample data, first, the preset time period is determined in the electric energy meter operation database according to the time index, and the preset time period is a continuous time interval in which the electric energy meter operation state is stable and no fault alarm record exists; then, the operation data in the time interval is screened, and the time segments with metering abnormality marks, communication abnormality marks or manual maintenance records are removed, and the operation data confirmed to be in a normal metering state by the operation record is retained; after the screening is completed, the corresponding electrical operation parameter data, chip real-time temperature feature data and discrete state code sequence and state transition record generated therefrom are extracted from the normal operation time interval to form a state transition sample sequence arranged in time sequence; and finally, the state transition sample sequence is taken as the feature learning sample data.

[0045] When the feature learning sample data is input into the state transition behavior model for behavior feature reverse learning processing, the state transition behavior model includes a state transition feature extraction framework, and the state transition feature extraction framework includes a state feature extraction layer and a transition feature extraction layer; the feature learning sample data is sequentially input into the state feature extraction layer in time sequence to extract single-time state features, and then input into the transition feature extraction layer to extract transition features between adjacent states, and the feature weights are optimized and solved based on the maximum entropy principle to obtain optimal state transition feature representation; and finally, the optimal state transition feature representation is stored as a standard state transition feature.

[0046] Further, in the method provided by the application embodiment, the feature learning sample data is input into the state transition behavior model for behavior feature reverse learning processing to extract a standard state transition feature, and the method further includes: The state transition behavior model includes a state transition feature extraction framework, and the state transition feature extraction framework includes a state feature extraction layer and a transition feature extraction layer; the feature learning sample data is sequentially input into the state feature extraction layer and the transition feature extraction layer, and the feature weight distribution is optimized through the maximum entropy principle to obtain optimal state transition feature representation; and the optimal state transition feature representation is stored as a standard state transition feature.

[0047] In the embodiment of the present application, the state transition behavior model comprises a state transition feature extraction framework, which contains a state feature extraction layer and a transition feature extraction layer. When performing reverse learning of behavior features, first, the feature learning sample data is input into the state feature extraction layer in chronological order. The feature learning sample data is state space basic data in a continuous time period and its corresponding discrete state encoding sequence. For each sampling time, the state space basic data can be represented as a 4-dimensional state feature vector [x1, x2, x3, x4] containing voltage, current, power factor and real-time temperature feature data of the chip. In the state feature extraction layer, an optimized weight parameter w1, w2, w3, w4 is set for each dimension of the feature, and the weights can be initially set as the same constant, for example, 1. Then, the state score value S = w1x1 + w2x2 + w3x3 + w4x4 of the time is calculated, so as to compress the original 4-dimensional state feature vector into one state score value, which is used to represent the overall running state intensity at the time.

[0048] Then, the state score values of two consecutive sampling times are input into the transition feature extraction layer. For time t and time t+1, first, the state change amount ΔS = S(t+1) - S(t) is calculated, which is used to represent the state change amplitude; then, a transition feature vector [T1, T2] is constructed, where T1 is the state score value S(t) of the current time, and T2 is the state change amount ΔS. Next, the optimized weight parameters a1 and a2 are set for the transition feature vector, and the transition score value Z = a1T1 + a2T2 is calculated, which is used to represent the importance of the state transition. The above calculation process is repeated for the state transition records in all samples to obtain a sequence of transition score values.

[0049] In the weight optimization phase, the maximum entropy principle is used for feature weight distribution. First, the empirical average values of T1 and T2 in the feature learning sample data are calculated, that is, the sum of T1 in all samples is divided by the total number of samples to obtain the empirical expectation E1, and the sum of T2 in all samples is divided by the total number of samples to obtain the empirical expectation E2. Then, the transition score value Z of each state transition record is calculated based on the current weight parameters, and the exponential operation is performed on all Z values, and the sum of the exponential results is obtained to obtain a normalization denominator. The exponential value of each record is divided by the denominator to obtain the corresponding transition probability. Then, the average values of T1 and T2 predicted by the model are recalculated according to the transition probability to obtain the predicted expectation value. The predicted expectation value is compared with the empirical expectation E1 and E2. When the predicted expectation is less than the empirical expectation, the corresponding weight is increased, and when the predicted expectation is greater than the empirical expectation, the corresponding weight is decreased. The steps of calculating the transition score value, the transition probability, the predicted expectation value and the weight adjustment are repeatedly calculated until the difference between the predicted expectation value and the empirical expectation value is less than the preset error threshold, which indicates that the weight parameters converge.

[0050] After the weight parameters converge, the final w1, w2, w3, w4, a1, a2 are used to recalculate all state score values and transition score values, and the state change amount ΔS is calculated according to the state score values of adjacent sampling time points to form a state score value sequence, a state change amount sequence and a transition score value sequence, and the corresponding transition feature is represented as an optimal state transition feature; finally, the optimal state transition feature is stored as a standard state transition feature, wherein the standard state transition feature includes a state score value sequence, a state change amount sequence and a transition score value sequence, and is used to completely represent the state evolution process under the normal operating condition.

[0051] Step S400: input the collected real-time operating state parameters into the state transition behavior model to obtain actual state transition features.

[0052] In the embodiments of the present application, when the collected real-time operating state parameters are input into the state transition behavior model, the real-time operating state parameters are first preprocessed, including time synchronization processing, outlier elimination and minimum maximum normalization processing, to generate standardized state input data; then the standardized state input data is input into the state transition behavior model, state transition simulation processing is performed under the constraint of the state transition probability description matrix, and key feature data in the state transition process is recorded to obtain actual state transition features.

[0053] Further, in the method provided by the embodiments of the present application, the collected real-time operating state parameters are input into the state transition behavior model to obtain actual state transition features, and the method further includes: The real-time operating state parameters are preprocessed to obtain standardized state input data; the standardized state input data is input into the state transition behavior model to perform state transition simulation processing, and key feature data in the state transition process is recorded to obtain actual state transition features.

[0054] In the embodiment of the present application, when the real-time running state parameters are preprocessed, first, time synchronization processing is performed on the collected real-time running state parameters. The time synchronization processing refers to aligning the electrical running parameters and the chip real-time temperature characteristic data according to the time stamp based on a unified sampling period, so that the parameters of each dimension at the same sampling time form a corresponding relationship. Then, abnormal value detection processing is performed. The adjacent difference value of the same dimension parameter at the continuous sampling time is calculated. When the absolute value of the difference value exceeds the preset abnormal threshold, it is determined as abnormal data, and the mean value of the adjacent effective data is used for replacement, so as to eliminate the instantaneous noise interference. After the abnormal value processing is completed, minimum maximum normalization processing is performed on each dimension parameter. According to the minimum value and the maximum value of the corresponding parameter in the historical normal running data, the current real-time running state parameter is linearly mapped to the interval of 0 to 1, so that different dimension parameters are in a unified numerical range. Finally, the multi-dimensional parameters after the time synchronization processing, the abnormal value detection processing and the minimum maximum normalization processing are combined to form standardized state input data.

[0055] Next, the standardized state input data is input into the state transition behavior model. State transition simulation processing is performed inside the state transition behavior model. First, the transition probability distribution corresponding to the current discrete state code is read based on the state transition probability description matrix, and the transition probability from the current state to each possible next state is calculated. Then, according to the transition probability distribution, a continuous state transition path is generated by probability sampling. The state transition path is composed of a plurality of sequentially evolved discrete state codes. During the state transition simulation processing, the key characteristic data in the state transition process is recorded in real time, including the state score value, the state change amount and the transition score value. Finally, the key characteristic data is summarized to form the actual state transition feature.

[0056] Further, the method provided by the application embodiment comprises the following steps: Based on the state transition probability description matrix in the state transition behavior model, the transition probability distribution from the current state to the next state is calculated. According to the transition probability distribution, a state transition path is simulated and generated. The key state features and transition features in the state transition path are extracted as the actual state transition features.

[0057] In the embodiment of the present application, when the transition probability distribution from the current state to the next state is calculated based on the state transition probability description matrix in the state transition behavior model, first, the current discrete state code is determined according to the standardized state input data, then the row corresponding to the current discrete state code in the state transition probability description matrix is found, and all the values in the row are read out in turn; each value in the row represents the probability value of transitioning from the current discrete state code to a certain next discrete state code; in order to ensure the validity of the probability, first, the sum of all the values in the row is obtained, if the sum is not equal to 1, then each value in the row is divided by the sum, so that the sum of all the corrected values is equal to 1; the corrected set of values is the transition probability distribution from the current state to each possible next state.

[0058] When the state transition path is simulated and generated according to the transition probability distribution, first, the transition probability distribution is sequentially accumulated to form a cumulative probability sequence, for example, if the three probabilities are 0.2, 0.5, and 0.3, the cumulative probability sequence is 0.2, 0.7, and 1.0; then a random number between 0 and 1 is generated, for example, 0.65, then the random number is compared with the cumulative probability sequence one by one, if the random number is greater than the previous cumulative value and less than or equal to the current cumulative value, the next discrete state code corresponding to the current cumulative value is selected as the new state; for example, 0.65 is greater than 0.2 and less than or equal to 0.7, the second state is selected; after the new discrete state code is determined, it is used as the new current discrete state code, and the steps of reading the probability distribution, calculating the cumulative probability sequence, and generating the random number are repeated, and the process is continuously performed according to the preset number of transitions, so that the state transition path formed by a plurality of discrete state codes connected in turn is obtained.

[0059] When the key state features and transition features in the state transition path are extracted as the actual state transition features, first, the standardized state input data corresponding to each discrete state code in the state transition path is found, and the state score value is calculated according to the determined weight parameter, that is, the state score value at this time is obtained by summing the product of each dimension of the standardized state input data and the corresponding weight, and the state score sequence is recorded in time sequence; then the state change amount is obtained by subtracting the adjacent state score values, and the transition score value is obtained by linearly weighting the current state score value and the state change amount according to the determined weight parameter; the state score sequence, the state change amount sequence, and the transition score sequence obtained in the entire state transition path are sorted and combined to form the actual state transition features.

[0060] Step S500: compare the difference between the actual state transition features and the standard state transition features, and when the difference exceeds the preset difference threshold, generate an electric energy meter health state prompt information.

[0061] In the embodiment of the present application, when comparing the difference degree of the actual state transition feature and the standard state transition feature, the similarity index of the actual state transition feature and the standard state transition feature is calculated. First, the actual state transition feature and the standard state transition feature are aligned, and both are arranged into the same length of feature vector form. The feature vector is obtained by splicing the state score value sequence, the state change amount sequence and the transition score value sequence in a predetermined order. When the lengths of the two sequences are inconsistent, the same length of data segment in the same time length is obtained by using the fixed window interception method, or the shorter sequence is padded by using the linear interpolation method, so that the two sequences correspond to each other at the same sampling time. After alignment, the similarity index is calculated by using the Euclidean distance. The difference value of each dimension of the feature vector is calculated and squared, and then the sum of all squared values is obtained and the square root is taken to obtain the distance value D. The distance value D is taken as the quantization result of the difference degree, wherein the larger the distance value D is, the higher the difference degree between the actual state transition feature and the standard state transition feature is.

[0062] When comparing and analyzing the similarity index and the preset difference threshold, first, the preset difference threshold is read. The preset difference threshold is calculated from the standard state transition feature of the historical normal running data. Specifically, the distance value between the standard state transition feature and the corresponding actual state transition feature is calculated on the historical normal running data, and the mean and the standard deviation of the distance value are taken. The preset difference threshold is determined in the manner of mean plus preset multiple standard deviation. Then, the distance value D calculated at present is compared with the preset difference threshold. When the distance value D is less than or equal to the preset difference threshold, it is determined that the difference degree does not exceed the allowable range. When the distance value D is greater than the preset difference threshold, it is determined that the difference degree exceeds the allowable range.

[0063] When generating the electric energy meter health state prompt information according to the comparison and analysis result, first, the normal health state prompt information is generated in the case that the difference degree does not exceed the allowable range. The electric energy meter health state is marked as normal, and the corresponding distance value D is recorded as the health evaluation result. In the case that the difference degree exceeds the allowable range, the abnormal health state prompt information is generated. The electric energy meter health state is marked as risk, and the warning mark is output. The distance value D and the amplitude of exceeding the preset difference threshold are recorded as the risk degree basis. Finally, the electric energy meter health state prompt information is output to the monitoring platform or the local display interface, which is used to prompt the deviation of the current running state of the electric energy meter relative to the standard state transition feature.

[0064] In the embodiment of the present application, as described above, the embodiment of the present application has at least the following technical effects: The application collects electrical operation parameters and temperature monitoring parameters of an electric energy meter in real time, the temperature monitoring parameters include chip temperature related data and environmental temperature data; real-time temperature feature data of a chip is extracted based on the chip temperature related data, the electrical operation parameters are combined to construct an electric energy meter operation state space, and a state transition behavior model is established based on the electric energy meter operation state space; the state transition behavior model is subjected to feature learning through historical normal operation data to extract standard state transition features; real-time operation state parameters collected are input into the state transition behavior model to obtain actual state transition features; the difference degree between the actual state transition features and the standard state transition features is compared, and when the difference degree exceeds a preset difference threshold, electric energy meter health state prompt information is generated. The application solves the technical problem that it is difficult to accurately and early identify an implicit fault of an electric energy meter in the prior art, and through construction of an operation state space and a state transition behavior model and comparison of standard and actual state transition features, the technical effect of accurate assessment of an electric energy meter health state and early warning of an implicit fault is achieved.

[0065] In the second embodiment, based on the same inventive concept as the electric energy meter health assessment method based on an operation state in the foregoing embodiments, as shown in the accompanying drawings, the application provides an electric energy meter health assessment system based on an operation state, and the system and method embodiments in the application are based on the same inventive concept. The system includes: Figure 2 A data collection module 11 is configured to collect electrical operation parameters and temperature monitoring parameters of an electric energy meter in real time, the temperature monitoring parameters include chip temperature related data and environmental temperature data; a model establishment module 12 is configured to extract real-time temperature feature data of a chip based on the chip temperature related data, combine the electrical operation parameters to construct an electric energy meter operation state space, and establish a state transition behavior model based on the electric energy meter operation state space; a feature learning module 13 is configured to perform feature learning on the state transition behavior model through historical normal operation data to extract standard state transition features; a transition feature acquisition module 14 is configured to input real-time operation state parameters collected into the state transition behavior model to obtain actual state transition features; and an information generation module 15 is configured to compare the difference degree between the actual state transition features and the standard state transition features, and when the difference degree exceeds a preset difference threshold, generate electric energy meter health state prompt information.

[0066] Further, the system is also configured to implement the following functions: ​The chip temperature-related data is subjected to temperature-sensitive feature extraction processing to obtain primary temperature feature data; the primary temperature feature data is subjected to temperature state calculation processing based on circuit board heat conduction characteristic data to obtain intermediate temperature feature data; and the intermediate temperature feature data is subjected to aging feature extraction processing according to a temperature cumulative effect analysis model to obtain chip real-time temperature feature data.

[0067] Further, the system is further used to implement the following functions: The electrical operation parameters and the chip real-time temperature feature data are subjected to parameter fusion processing to obtain state space basic data; a state change mode is defined based on the state space basic data to obtain state space change description data; and state transition law analysis is performed on the state space basic data and the state space change description data according to device historical operation data to construct an electric energy meter operation state space.

[0068] Further, the system is further used to implement the following functions: The state space basic data and the state space change description data are obtained from the electric energy meter operation state space; a state transition probability description matrix containing state transition conditions and probability values is constructed based on the state space basic data and the state space change description data; and a state transition behavior model is established according to the state transition probability description matrix.

[0069] Further, the system is further used to implement the following functions: Historical operation state transition sequence data is collected; transition frequencies and transition conditions between states are counted based on the historical operation state transition sequence data; and a state transition probability description matrix is constructed according to the transition frequencies and the transition conditions, and a probability updating rule is set.

[0070] Further, the system is further used to implement the following functions: Historical normal operation data in a preset time period is selected as feature learning sample data; the feature learning sample data is input into the state transition behavior model for behavior feature reverse learning processing to extract standard state transition features.

[0071] Further, the system is further used to implement the following functions: The state transition behavior model comprises a state transition feature extraction framework, and the state transition feature extraction framework comprises a state feature extraction layer and a transition feature extraction layer; the feature learning sample data is input into the state feature extraction layer and the transition feature extraction layer in sequence, and feature weight distribution is optimized through a maximum entropy principle to obtain optimal state transition feature representation; and the optimal state transition feature representation is stored as a standard state transition feature.

[0072] Further, the system is also used to realize the following functions: The real-time running state parameters are preprocessed to obtain standardized state input data; the standardized state input data are input into the state transition behavior model, state transition simulation processing is performed, key feature data in the state transition process are recorded, and actual state transition features are obtained.

[0073] Further, the system is also used to realize the following functions: Based on the state transition probability description matrix in the state transition behavior model, the transition probability distribution from the current state to the next state is calculated; the state transition path is simulated and generated according to the transition probability distribution; and the key state features and transition features in the state transition path are extracted as actual state transition features.

[0074] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0075] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the present application, and any modification, equivalent change and modification made on the basis of the technical essence of the present application to the above embodiments are still within the scope of the present application.

Claims

1. A method for assessing the health of electricity meters based on their operating status, characterized in that, The method includes: The electrical operating parameters and temperature monitoring parameters of the electricity meter are collected in real time. The temperature monitoring parameters include chip temperature-related data and ambient temperature data. Based on the chip temperature-related data, real-time temperature feature data of the chip is extracted. Combined with the electrical operating parameters, the operating state space of the electricity meter is constructed. Based on the operating state space of the electricity meter, a state transition behavior model is established. The state transition behavior model is subjected to feature learning using historical normal operation data to extract standard state transition features; The collected real-time operating status parameters are input into the state transition behavior model to obtain the actual state transition characteristics; The difference between the actual state transition characteristics and the standard state transition characteristics is compared. When the difference exceeds a preset difference threshold, a health status prompt message for the electricity meter is generated.

2. The method for assessing the health of an energy meter based on its operating status as described in claim 1, characterized in that, Based on the chip temperature-related data, real-time temperature feature data of the chip is extracted, including: Temperature-sensitive feature extraction processing is performed on the chip temperature-related data to obtain primary temperature feature data; Based on the thermal conductivity data of the circuit board, the temperature state is calculated from the primary temperature characteristic data to obtain the intermediate temperature characteristic data. Based on the temperature accumulation effect analysis model, the intermediate temperature feature data is processed by aging feature extraction to obtain the chip's real-time temperature feature data.

3. The method for assessing the health of an energy meter based on its operating status as described in claim 2, characterized in that, Construct the operating state space of the electricity meter, including: The electrical operating parameters and the real-time temperature characteristic data of the chip are fused together to obtain the basic state space data. Based on the aforementioned state space basic data, state change modes are defined to obtain state space change description data; Based on the historical operating data of the equipment, the state transition law is analyzed on the basic data of the state space and the state space change description data to construct the operating state space of the electricity meter.

4. The method for assessing the health of an energy meter based on its operating status as described in claim 1, characterized in that, Based on the operating state space of the electricity meter, a state transition behavior model is established, including: Obtain basic state space data and state space change description data from the operating state space of the energy meter; Based on the state space basic data and state space change description data, a state transition probability description matrix containing state transition conditions and probability values ​​is constructed. A state transition behavior model is established based on the state transition probability description matrix.

5. The method for assessing the health of an energy meter based on its operating status as described in claim 4, characterized in that, Constructing the state transition probability description matrix also includes: Collect historical operational status transition sequence data; Based on the historical operating state transition sequence data, the transition frequency and transition conditions between each state are statistically analyzed; Based on the transition frequency and transition conditions, a state transition probability description matrix is ​​constructed, and probability update rules are set.

6. The method for assessing the health of an energy meter based on its operating status as described in claim 1, characterized in that, The state transition behavior model is subjected to feature learning using historical normal operation data to extract standard state transition features, including: Historical normal operation data within a preset time period is selected as feature learning sample data; The feature learning sample data is input into the state transition behavior model to perform inverse learning of the behavior features and extract standard state transition features.

7. The method for assessing the health of an energy meter based on its operating status as described in claim 6, characterized in that, The feature learning sample data is input into the state transition behavior model to perform inverse learning of the behavior features and extract standard state transition features, including: The state transition behavior model includes a state transition feature extraction framework, which comprises a state feature extraction layer and a transition feature extraction layer. The feature learning sample data is sequentially input into the state feature extraction layer and the transition feature extraction layer, and the feature weight allocation is optimized by the maximum entropy principle to obtain the optimal state transition feature representation. The optimal state transition feature is stored as a standard state transition feature.

8. The method for assessing the health of an energy meter based on its operating status as described in claim 1, characterized in that, The collected real-time operating status parameters are input into the state transition behavior model to obtain the actual state transition characteristics, including: The real-time operating status parameters are preprocessed to obtain standardized status input data; The standardized state input data is input into the state transition behavior model, and state transition simulation processing is performed. Key feature data during the state transition process are recorded to obtain the actual state transition features.

9. The method for assessing the health of an energy meter based on its operating status as described in claim 8, characterized in that, The standardized state input data is input into the state transition behavior model, and state transition simulation processing is performed. Key feature data during the state transition process are recorded to obtain the actual state transition features, including: Based on the state transition probability description matrix in the state transition behavior model, calculate the transition probability distribution from the current state to the next state; Based on the aforementioned transition probability distribution, state transition paths are simulated and generated; Extract the key state features and transition features from the state transition path as the actual state transition features.

10. A health assessment system for electricity meters based on operational status, characterized in that, The system is used to perform the energy meter health assessment method based on operating status as described in any one of claims 1-9, and the system comprises: The data acquisition module is used to collect electrical operating parameters and temperature monitoring parameters of the electricity meter in real time. The temperature monitoring parameters include chip temperature-related data and ambient temperature data. The model building module is used to extract real-time temperature feature data of the chip based on the chip temperature-related data, combine it with the electrical operating parameters, construct the operating state space of the electricity meter, and establish a state transition behavior model based on the operating state space of the electricity meter. The feature learning module is used to learn features from the state transition behavior model using historical normal operation data and extract standard state transition features. The transition feature acquisition module is used to input the collected real-time running status parameters into the state transition behavior model to obtain the actual state transition features; The information generation module is used to compare the degree of difference between the actual state transition characteristics and the standard state transition characteristics. When the degree of difference exceeds a preset difference threshold, it generates a health status prompt message for the electricity meter.

Images