Electric energy meter with power data jump anomaly intelligent prediction function

Abstract

The present application relates to the technical field of data processing, and more particularly to an electric energy meter with power data jump abnormality intelligent prediction function, which comprises a processor and a memory, and the processor executes the computer program of the memory to realize the following steps: based on the fault physical mechanism, the coupling relationship between the voltage abnormality score and the clock deviation score of the target electric energy meter at each time within a preset time period up to the current time, and the current abnormality score at each time are analyzed, the jump abnormality coefficient at each time is obtained, the jump abnormality coefficient is optimized according to the historical jump abnormality risk level before each time and the evolution speed of the jump abnormality coefficient at each time, and the jump abnormality optimization coefficient at each time is obtained; the jump abnormality probability of the target electric energy meter at a future time is predicted according to the jump abnormality optimization coefficient at each time, and the source early warning and predictive maintenance of the power data jump abnormality are realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an energy meter with intelligent prediction function for abnormal fluctuations in power data. Background Technology

[0002] In the context of the rapid development of smart grids and digital twin systems, the accuracy and stability of electricity meters, as the data sensing endpoints of the power grid, are the cornerstone of condition assessment, line loss analysis, and transaction settlement. However, in actual operation, battery degradation and clock deviation in electricity meters may work together to cause systemic errors, leading to abnormal fluctuations in power data and seriously threatening data quality.

[0003] When a data-logging anomaly occurs in an existing electricity meter, it can typically only mark and store the abnormal results (such as voltage or current loss) after the anomaly occurs. This is essentially a post-event recording and lacks in-depth integration and trend analysis of continuous, multi-dimensional electrical characteristics and physical status data before the anomaly occurs. It is difficult to quantitatively assess the meter's own health degradation trajectory, making it impossible for the operation and maintenance side to distinguish between occasional interference and inevitable fault precursors, and unable to provide early warnings before the anomaly occurs. This blurs the line between predictive maintenance and post-event repair, which not only reduces operation and maintenance efficiency but also makes it difficult to effectively assess and extend the reliable operating life of the electricity meter in complex field environments.

[0004] Therefore, how to provide early warnings of abnormal fluctuations in the power data of electricity meters has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an energy meter with intelligent prediction function for abnormal power data fluctuations, in order to solve the problem of how to provide early warning for abnormal power data fluctuations in energy meters.

[0006] This invention provides an energy meter with intelligent prediction function for abnormal power data fluctuations, including a memory, a processor, and a computer program stored in the memory and running on the processor. The processor executes the computer program to perform the following steps:

[0007] Based on the rated current, battery critical voltage, and maximum allowable clock deviation of the target energy meter, the A-phase current, battery voltage, and clock deviation between the built-in clock and the main clock of the target energy meter are piecewise linearly normalized at each moment within the preset time period up to the current moment to obtain the current abnormality score, voltage abnormality score, and clock deviation score of the target energy meter at each moment.

[0008] Based on the physical mechanism of the fault, the coupling relationship between the voltage anomaly fraction and the clock deviation fraction at each moment and the current anomaly fraction at each moment are analyzed to obtain the jump anomaly coefficient of the target energy meter at each moment. According to the historical jump anomaly risk level of the target energy meter before each moment and the evolution rate of the jump anomaly coefficient at each moment, the jump anomaly coefficient at each moment is optimized to obtain the jump anomaly optimization coefficient of the target energy meter at each moment within the preset time period.

[0009] Based on the anomaly optimization coefficient at each time point, the probability of the target energy meter experiencing anomalies at future time points is predicted, and the anomaly probability at future time points is obtained. Based on the anomaly probability, an anomaly warning is issued for the power data anomalies of the target energy meter at future time points.

[0010] Preferably, the step of analyzing the coupling relationship between the voltage anomaly score and the clock deviation score at each time moment, as well as the current anomaly score at each time moment, to obtain the jump anomaly coefficient of the target energy meter at each time moment, includes:

[0011] For any given moment, calculate the product between the voltage anomaly fraction and the clock deviation fraction at that given moment to obtain the first anomaly level of the target energy meter at that given moment;

[0012] Calculate the reciprocal of the sum of the current anomaly fraction at any given time and a preset constant, and use the negative of the product of the reciprocal and a preset sensitivity coefficient as the independent variable of the natural exponential function to obtain the second anomaly degree of the target energy meter at any given time.

[0013] The first anomaly level and the second anomaly level are weighted and summed to obtain the jump anomaly coefficient of the target energy meter at any given time.

[0014] Preferably, the step of optimizing the jump anomaly coefficient at each moment based on the historical jump anomaly risk level of the target energy meter before each moment and the evolution rate of the jump anomaly coefficient at each moment, to obtain the optimized jump anomaly coefficient of the target energy meter at each moment within the preset time period, includes:

[0015] Based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step, the jump anomaly risk momentum of the target energy meter at each time step is obtained.

[0016] For any given moment, the product of the risk momentum of the jump anomaly at that moment and the preset sensitivity coefficient is used as the independent variable of the hyperbolic tangent function to obtain the optimization vector for optimizing the jump anomaly coefficient at that moment. The product between the preset optimization degree and the optimization vector is calculated, and the sum of the constant 1 and the product is used as the optimization factor for optimizing the jump anomaly coefficient at that moment.

[0017] Calculate the product between the jump anomaly coefficient at any given time and the optimization factor to obtain the jump anomaly optimization coefficient of the target energy meter at any given time.

[0018] Preferably, obtaining the jump anomaly risk momentum of the target energy meter at each time step based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step includes:

[0019] For any given moment, based on the jump anomaly coefficient of the target energy meter at each moment between the start of operation and that moment, the historical jump anomaly risk quality of the target energy meter up to that moment is obtained using an exponential weighted average algorithm;

[0020] Subtracting the jump anomaly coefficient of the previous moment from the jump anomaly coefficient at any given moment yields the instantaneous change rate of the target energy meter at any given moment.

[0021] Calculate the product between the historical jump anomaly risk quality and the instantaneous change rate of the jump anomaly to obtain the risk characteristic value of the target energy meter at any given moment;

[0022] During the historical normal operation of the target energy meter, the risk quality of the jump anomaly and the instantaneous change rate of the jump anomaly at each historical normal moment within the historical normal operation period of a preset duration are obtained. The target energy meter does not experience any power data jump anomaly during the historical normal operation period, and the A-phase current, battery voltage and clock deviation of the target energy meter at each historical normal moment do not exceed the rated current, battery critical voltage and maximum allowable clock deviation of the target energy meter.

[0023] Calculate the standard deviation of the risk quality of the jump anomaly at all historical normal times, denoted as the risk quality standard deviation. Calculate the standard deviation of the instantaneous rate of change of the jump anomaly at all historical normal times, denoted as the instantaneous rate of change standard deviation. Calculate the product between the risk quality standard deviation and the instantaneous rate of change standard deviation. Use a preset constant and the product as the denominator, and use the risk characteristic value of the target energy meter at any given time as the numerator to obtain the jump anomaly risk momentum of the target energy meter at any given time. The jump anomaly risk momentum is a vector.

[0024] Preferably, the step of predicting the probability of a jump anomaly occurring in the target energy meter at a future time based on the jump anomaly optimization coefficient at each time moment, to obtain the jump anomaly probability at the future time moment, includes:

[0025] Based on the time interval between each moment of the target energy meter within the preset time period and the current moment, the exponential decay weight of each moment within the preset time period is obtained using an exponential decay function with a preset decay factor as the base.

[0026] Based on the exponential decay weight of each moment within the preset time period, the jump anomaly optimization coefficients of the target energy meter at each moment within the preset time period are weighted and summed to obtain the summation result. The summation result and the preset bias value are used as the independent variable of the sigmoid activation function to obtain the jump anomaly probability of the target energy meter at future moments.

[0027] Preferably, the step of providing an anomaly warning for abnormal power data fluctuations of the target energy meter at future times based on the anomaly probability includes:

[0028] The probability of a jump anomaly in the target energy meter at a historical moment when no jump anomaly occurred within a preset time window before the future moment is obtained and recorded as the historical jump anomaly probability. The percentile of the jump anomaly probability at the future moment is obtained among all historical jump anomaly probabilities. If the percentile is greater than a preset percentile threshold, an anomaly warning is issued for the power data jump anomaly of the target energy meter at the future moment.

[0029] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0030] This invention first analyzes the coupling relationship between voltage anomaly fraction and clock deviation fraction at each time step, as well as current anomaly fraction, based on the physical mechanism of the fault, to obtain the jump anomaly coefficient of the target energy meter at each time step, thus achieving direct and interpretable quantification of the jump risk intensity of the energy meter. Then, based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step, the jump anomaly coefficient at each time step is optimized to obtain the jump anomaly optimization coefficient. This enables more sensitive identification of continuous and accelerating jump anomaly change trends, deepening the static jump anomaly coefficient into a dynamic trend criterion, providing key input for the final accurate prediction. Finally, based on the jump anomaly optimization coefficient at each time step, lightweight time-series prediction is performed on the jump anomaly probability of the target energy meter at future time steps, realizing source early warning and predictive maintenance of power data jump anomalies. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0032] Figure 1 This is a flowchart of a method for implementing an energy meter with intelligent prediction function for power data jump anomalies, provided in Embodiment 1 of the present invention. Detailed Implementation

[0033] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0034] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0035] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0036] See Figure 1 This is a flowchart illustrating a method for implementing an energy meter with intelligent prediction function for power data anomalies, as provided in Embodiment 1 of the present invention. Figure 1 As shown, the method may include:

[0037] Step S101: Based on the rated current of the target energy meter, the critical voltage of the battery, and the maximum allowable clock deviation, the A-phase current, battery voltage, and clock deviation between the built-in clock and the main clock of the target energy meter at each moment within the preset time period up to the current moment are processed by piecewise linear normalization to obtain the current abnormality score, voltage abnormality score, and clock deviation score of the target energy meter at each moment.

[0038] Battery and clock coordination failure is one of the root causes of abnormal power data jumps in electricity meters. At the same time, abnormal current is the most direct physical trigger point for data jumps. Battery failure causes the clock to run slow. Suppose that when the real time reaches 0 o'clock, the internal clock of the electricity meter may only be 23:30, and the freeze task is not triggered, resulting in no frozen data for that day. The main station system has a blank value. The next day, the main station system tries to re-read yesterday's data, but due to the clock deviation, the "yesterday's electricity consumption" reported by the electricity meter actually includes the sum of the data from the last period of yesterday and the first period of today. On the electricity consumption curve, the electricity consumption is zero (or extremely low) on a certain day, followed by an abnormally high peak the next day, forming a cliff-like jump. Therefore, in this embodiment of the invention, the risk level of data jump anomalies in the electricity meter can be characterized by the A-phase current of the electricity meter, the battery voltage, and the clock deviation between the electricity meter's built-in clock and the master clock (the electricity meter has only one hardware clock, usually provided by the RTC chip or the RTC module inside the MCU, which is responsible for providing a time reference for metering, freezing, and rate switching. The electricity meter receives time synchronization commands from the master station or concentrator of the electricity information collection system. At this time, the master station system or concentrator acts as the "master clock," and the clock inside the electricity meter is passively calibrated as the "slave clock"). This allows for risk prediction of future jump anomalies in the electricity meter, thereby achieving source warning and predictive maintenance of electricity meter power data jump anomalies.

[0039] Any energy meter awaiting warning is designated as the target energy meter. From the moment the target energy meter is put into operation, the instantaneous sampled values ​​of three key physical quantities—the effective value of phase A current, battery voltage, and clock deviation from the master station—are read in real time at a fixed sampling period through the metering chip, power management unit, and clock module of the target energy meter. In this embodiment, the sampling period is set to 1 minute, but this is not limited and can be set by the implementer according to the specific scenario. That is, the instantaneous sampled values ​​of phase A current, battery voltage, and clock deviation of the target energy meter are read every minute. Furthermore, the read instantaneous sampled values ​​of phase A current, battery voltage, and clock deviation are preprocessed to transform them into uniform and dimensionless data, obtaining the current anomaly score, voltage anomaly score, and clock deviation score at each moment, which facilitates subsequent analysis and calculation.

[0040] Among them, data preprocessing includes, but is not limited to: (1) performing rationality verification (such as range check) on the collected instantaneous sampled values, and filling the single-point data missing caused by instantaneous communication interruption with the nearest point linear interpolation method to ensure the continuity of the data sequence; (2) applying moving average filtering or first-order low-pass filtering to filter the verified and repaired instantaneous sampled values ​​to suppress high-frequency random noise and short-time pulse interference, improve the signal-to-noise ratio of the data, and form a set of cleaner physical quantities that are more stable and better reflect the real trend; (3) according to the rated current of the target energy meter, the critical voltage of the battery and the maximum allowable clock deviation, normalizing the filtered data through piecewise linear functions, converting it into dimensionless normalized values ​​in the range of 0-1, wherein the linear mapping relationship between the effective value of the A-phase current and the battery voltage is: the closer the effective value of the A-phase current and the battery voltage are to the rated current and the critical voltage of the battery after filtering, the smaller their normalized values, indicating that they are more normal and the possibility of abnormality is smaller. The linear mapping of clock deviation is as follows: the smaller the filtered clock deviation, the smaller the normalized value, indicating that it is more normal and the possibility of anomaly is smaller. For piecewise linear functions, in this embodiment of the invention, the piecewise linear function is set to a two-segment form, and the mapping function of each segment is the maximum and minimum normalized function. There is no restriction here. The implementer can set the number of segments of the piecewise linear function and the mapping function of each segment according to the specific scenario. (4) The normalized value corresponding to the effective value of the A-phase current is recorded as the current anomaly score, the normalized value corresponding to the battery voltage is recorded as the voltage anomaly score, and the normalized value corresponding to the clock deviation is recorded as the clock deviation score. Data preprocessing is existing technology and will not be described in detail here.

[0041] Since the present invention provides an instantaneous and rapid prediction of the risk of meter jump anomalies in the future, the computational load should not be too large, i.e., the amount of historical data referenced should not be too much. Therefore, in the embodiments of the present invention, a recent short window is used as the basis to fit the recent development of jump anomalies in order to achieve intelligent early warning of jump anomalies in advance, and to meet the real-time requirements of the meter. The current anomaly score, voltage anomaly score, and clock deviation score at each moment within 10 minutes up to the current moment are used as historical data for reference to predict the jump anomalies of the meter in the future. That is, the preset time period is set to 10 minutes, which is not limited here, and the implementer can set it according to the specific scenario.

[0042] Step S102: Based on the physical mechanism of the fault, analyze the coupling relationship between the voltage anomaly score and the clock deviation score at each time point, as well as the current anomaly score at each time point, to obtain the jump anomaly coefficient of the target energy meter at each time point. Based on the historical jump anomaly risk level of the target energy meter before each time point and the evolution rate of the jump anomaly coefficient at each time point, optimize the jump anomaly coefficient at each time point to obtain the optimized jump anomaly coefficient of the target energy meter at each time point within the preset time period.

[0043] Since battery degradation and clock skew in electricity meters, which work together to cause systematic errors, are one of the root causes of abnormal power data jumps, and current anomalies are the most direct physical trigger for power data jumps, this invention first analyzes the coupling relationship between voltage anomaly fraction and clock skew fraction at each moment, as well as current anomaly fraction at each moment, based on the physical causes of power data jump faults in electricity meters. This allows for the acquisition of the jump anomaly coefficient of the target electricity meter at each moment, thereby achieving a direct and interpretable quantification of the risk intensity of power data jump anomalies at each moment.

[0044] Taking time t as an example, the specific method for obtaining the jump anomaly coefficient of the target energy meter at time t is as follows:

[0045] Considering that the voltage of a single battery is low, the clock can maintain basic calibration; a slight deviation in the individual clock may just be network latency, but if both deteriorate at the same time, it means that the core foundation of the device is collapsing, leading to an increase in risk intensity and an increased probability of jump anomalies. Therefore, the product between the voltage anomaly score and the clock deviation score at time t is calculated to obtain the first anomaly level of the target energy meter at time t.

[0046] Since current anomaly is the most direct physical trigger point for power data jumps, when the A-phase current is normal (close to the rated current), the smaller the current anomaly fraction, the closer it is to 0. When the A-phase current is abnormal (far below or far above the rated current), the larger the current anomaly fraction, the closer it is to 1. Therefore, the reciprocal of the sum between the current anomaly fraction at time t and the preset constant is calculated. The opposite of the product between the reciprocal and the preset sensitivity coefficient is used as the independent variable of the natural exponential function to obtain the second anomaly degree of the target energy meter at time t.

[0047] The first anomaly level and the second anomaly level are weighted and summed to obtain the jump anomaly coefficient of the target energy meter at time t.

[0048] In one embodiment, the formula for calculating the jump anomaly coefficient of the target energy meter at time t is:

[0049]

[0050] in, This represents the jump anomaly coefficient of the target energy meter at time t. This represents the voltage anomaly fraction of the target energy meter at time t. This represents the clock deviation fraction of the target energy meter at time t, and k represents the preset sensitivity coefficient. This represents the fraction of current anomalies in the target energy meter at time t. This represents a preset constant. Indicates the first weight. Indicates the second weight. This represents the natural exponential function.

[0051] It should be noted that, The larger and The larger the value, the lower the battery voltage of the target energy meter at time t, and the greater the clock deviation. In this case, the greater the likelihood that the battery degradation and clock deviation will work together to cause systemic errors. The larger; The larger the value, the more the phase A current of the target energy meter deviates from the rated current at time t. This is very likely a precursor to a tripping event in the target energy meter, or a tripping event has already occurred. The larger the value. For the first weight... Second weight The settings are as follows: Since the coupling relationship between voltage anomaly fraction and clock deviation fraction, and current anomaly fraction belong to two different levels characterizing power data jump anomalies, the embodiments of the present invention are configured as follows: Regarding the setting of the preset sensitivity coefficient k, k is used to control the sensitivity of phase A current to abnormal power data fluctuations. Based on experimental statistics, k is set to 0.3. Regarding the preset constant... The settings, To ensure that the fraction is meaningful, the embodiments of the present invention are configured as follows: ,for , , k and There are no restrictions on the settings; implementers can configure them according to specific scenarios.

[0052] Similarly, obtain the jump anomaly coefficient of the target electricity meter at each moment within 10 minutes up to the current moment.

[0053] Considering that the anomaly coefficient of the target energy meter at each moment is essentially a snapshot, it can only quantify static risk and cannot distinguish whether it is an inevitable result of continuous deterioration, an inevitable result of continuous improvement, or a momentary fluctuation caused by occasional interference. Therefore, it is necessary to further optimize the anomaly coefficient at each moment based on the historical anomaly risk level of the target energy meter before each moment and the evolution rate of the anomaly coefficient at each moment. This will yield the optimized anomaly coefficient of the target energy meter at each moment, thus transforming the static anomaly coefficient into a dynamic trend criterion, keenly sensing the risk development trend, and effectively filtering momentary interference, providing key input for the final accurate prediction.

[0054] Taking time t as an example, the steps to obtain the optimization coefficient of the target energy meter at time t are as follows:

[0055] (1) Based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step, obtain the jump anomaly risk momentum of the target energy meter at each time step.

[0056] Specifically, based on the jump anomaly coefficient of the target energy meter at each time point from the start of operation to time t, the Exponential Weighted Average (EWMA) algorithm is used to obtain the historical jump anomaly risk quality of the target energy meter up to time t. This risk level is used to quantify the jump anomaly coefficient of the target energy meter up to time t, denoted as . ,Right now ,in, It represents the historical anomaly risk quality of the target electricity meter up to time t, which is also the exponentially weighted average (EWMA value) at time t. This represents the decay factor in the exponentially weighted average algorithm. This represents the exponentially weighted average (EWMA) value at time t-1. This represents the jump anomaly coefficient and attenuation factor of the target energy meter at time t. The value is set to 0.5; there is no restriction here, and implementers can set it according to the specific scenario. The larger the value, the higher the overall risk level of abnormal fluctuations in the target energy meter from the start of operation to time t, and vice versa. The smaller the value, the lower the overall risk level of abnormal fluctuations in the target energy meter from the start of operation to time t. The value is relatively small; the exponential weighted average algorithm is existing technology and will not be elaborated here.

[0057] Subtract the jump anomaly coefficient of the previous time (time t-1) from the jump anomaly coefficient of time t at time t to obtain the instantaneous change rate of the target energy meter at time t.

[0058] The risk characteristic value of the target energy meter at time t is obtained by calculating the product between the historical jump anomaly risk quality and the instantaneous change rate of the jump anomaly.

[0059] Since different electricity meters have different states, it is necessary to utilize the characteristics of the target electricity meter itself to standardize the risk characteristic value of the target electricity meter at time t, so as to obtain the jump anomaly risk momentum of the target electricity meter at time t. The specific implementation method is as follows: during the historical normal operation of the target electricity meter, the jump anomaly risk quality and the instantaneous change rate of the jump anomaly at each historical normal time within the historical normal operation period of a preset duration are obtained. The target electricity meter does not experience any power data jump anomalies during the historical normal operation period, and the A-phase current, battery voltage, and clock deviation of the target electricity meter at each historical normal time do not exceed the rated current of the target electricity meter, the critical voltage of the battery, and the maximum allowable clock deviation. The preset duration is set to 30 days, which is not limited here and can be set by the implementer according to the specific scenario.

[0060] Calculate the standard deviation of the risk quality of the jump anomaly at all historical normal times, denoted as the risk quality standard deviation. Calculate the standard deviation of the instantaneous change rate of the jump anomaly at all historical normal times, denoted as the instantaneous change rate standard deviation. Calculate the product between the risk quality standard deviation and the instantaneous change rate standard deviation. Use a preset constant and the product as the denominator, and use the risk characteristic value of the target energy meter at time t as the numerator to obtain the jump anomaly risk momentum of the target energy meter at time t. The jump anomaly risk momentum is a vector.

[0061] In one embodiment, the formula for calculating the risk momentum of the target energy meter at time t is:

[0062]

[0063] in, This represents the momentum of the target energy meter at time t, indicating the risk of anomaly. This represents the historical anomaly risk quality of the target electricity meter up to time t. This represents the jump anomaly coefficient of the target energy meter at time t. This represents the jump anomaly coefficient of the target energy meter at time t-1. Indicates the standard deviation of risk quality. Indicates the standard deviation of the instantaneous rate of change. This represents a preset constant.

[0064] It should be noted that, if , A positive number indicates that the abnormal fluctuation in the power data of the target energy meter at time t is worsening, and the abnormal fluctuation coefficient of the target energy meter at time t should be increased. The larger the value, the higher the overall risk level of abrupt changes before time t. To improve the sensitivity of identifying abrupt changes, the increase in the abrupt change coefficient at time t should be greater, i.e. The larger, the smaller, The smaller the value, the lower the overall risk level of the jump anomaly before time t, indicating that the target energy meter itself has a safe system with risk resistance. In this case, the response to instantaneous deterioration needs to be more conservative, meaning the increase in the jump anomaly coefficient at time t should be smaller. The smaller;

[0065] like , A negative value indicates that the abnormal fluctuations in the target energy meter's power data at time t have been alleviated, and the abnormal fluctuation coefficient of the target energy meter at time t should be appropriately reduced. The larger the value, the higher the overall risk level of power jump anomalies before time t. The target energy meter shows some relief in power jump anomalies at time t, and this mitigation trend is a significant positive signal. Therefore, this mitigation trend should be amplified; that is, the greater the reduction in the power jump anomaly coefficient at time t, the better. The larger, the smaller, The smaller the value, the lower the overall risk level of abrupt changes before time t, and the smaller its influence on the risk trend of abrupt changes. Therefore, the reduction in the abrupt change coefficient at time t should be smaller. The smaller;

[0066] like This indicates that the risk level of the target energy meter's jump anomaly at time t has not changed. That is, the jump anomaly coefficient at time t is not adjusted.

[0067] and This indicates the target energy meter's own state, used for... Perform standardization processing. For preset constants... The settings, To ensure that the fraction is meaningful, the embodiments of the present invention are configured as follows: There are no restrictions here; implementers can set them according to the specific scenario.

[0068] (2) Based on the risk momentum of the jump anomaly of the target energy meter at time t, the jump anomaly coefficient at time t is optimized to obtain the optimized jump anomaly coefficient of the target energy meter at time t.

[0069] Specifically: the product between the risk momentum of the jump anomaly at time t and the preset sensitivity coefficient is used as the independent variable of the hyperbolic tangent function to obtain the optimization vector for optimizing the jump anomaly coefficient at time t. The product between the preset optimization degree and the optimization vector is calculated, and the sum of the constant 1 and the product is used as the optimization factor for optimizing the jump anomaly coefficient at time t.

[0070] The product of the jump anomaly coefficient at time t and the optimization factor is calculated to obtain the jump anomaly optimization coefficient of the target energy meter at time t.

[0071] In one embodiment, the formula for calculating the optimization coefficient of the target energy meter at time t is:

[0072]

[0073] in, This represents the optimization coefficient for the jump anomaly of the target energy meter at time t. This represents the jump anomaly coefficient of the target energy meter at time t. Indicates the preset optimization level. This represents the preset sensitivity coefficient. This represents the momentum of the target energy meter at time t, indicating the risk of anomaly. This represents the hyperbolic tangent function.

[0074] It should be noted that, If the signal indicates that the abnormal fluctuation in the power data of the target energy meter at time t is worsening, then the abnormal fluctuation coefficient of the target energy meter at time t should be appropriately increased. The larger, The closer it is to 1, the better. The greater the increase, the more... The larger; When this occurs, it indicates that the abnormal fluctuations in the target energy meter's power data at time t have improved. Therefore, the abnormal fluctuation coefficient of the target energy meter at time t should be appropriately reduced. The larger, The closer it is to -1, the better. The greater the decrease, the more... The smaller; This indicates that the abnormal fluctuation in the power data of the target energy meter at time t has not changed. , That is, the jump anomaly coefficient at time t is not adjusted.

[0075] For the preset optimization level The settings, Used to control optimization vectors The maximum adjustment range for the jump anomaly coefficient is (0,1). In this embodiment of the invention, it is set... This indicates that the optimized vector can amplify or reduce the jump anomaly coefficient by a maximum of 30%, but there is no limit here; implementers can set it according to specific scenarios. (Regarding the preset sensitivity coefficient...) The settings, Used to control the sensitivity of the hyperbolic tangent function in the zero region, so that The curve is approximately linear. The value range of is 1-3, since the range of the hyperbolic tangent function is (-1, 1). The range of its value is [0, 1]. The value is approximately 0.964, which is close to saturation, so it is set to... There are no restrictions here; implementers can set them according to the specific scenario.

[0076] Thus, the optimization coefficient of the target energy meter at time t is obtained. Similarly, the optimization coefficient of the target energy meter at each time point within the 10 minutes up to the current time is obtained.

[0077] Step S103: Based on the jump anomaly optimization coefficient at each time moment, predict the probability of the target energy meter experiencing a jump anomaly at a future time moment to obtain the jump anomaly probability at a future time moment. Based on the jump anomaly probability, issue an anomaly warning for the target energy meter's power data jump anomaly at a future time moment.

[0078] After obtaining the optimized coefficients for the jump anomaly of the target energy meter at each time point within the 10 minutes up to the current time, a lightweight time-series prediction is performed on the probability of jump anomalies of the target energy meter at future time points based on the development of the optimized coefficients at each time point within the 10 minutes. This enables intelligent early warning of jump anomalies in advance. Specifically:

[0079] Based on the time interval between each moment of the target energy meter within the preset time period and the current moment, the exponential decay weight of each moment within the preset time period is obtained using an exponential decay function with a preset decay factor as the base.

[0080] Based on the exponential decay weight of each moment within the preset time period, the jump anomaly optimization coefficients of the target energy meter at each moment within the preset time period are weighted and summed to obtain the summation result. In this embodiment of the invention, the summation result is mapped to an intuitive jump anomaly probability between 0 and 1 using the sigmoid activation function. When the jump anomaly optimization coefficients of the target energy meter at each moment within the preset time period all infinitely approach 0 or equal to 0, it indicates that the target energy meter is in a healthy state within the preset time period and there is no risk of data jump anomalies. At this time, the basic jump anomaly probability is used as the sigmoid activation function. The output value of the id activation function is set to a basic jump anomaly probability of 0.1% in this embodiment of the invention. This is not a limitation and can be set by the implementer according to the specific scenario. That is, the output value of the sigmoid activation function is 0.001. Using the formula to deduce, the input value of the sigmoid activation function is approximately -6.9. Therefore, the input value -6.9 corresponding to the output value of the sigmoid activation function is recorded as the preset bias value. The summation result and the preset bias value are used as the independent variable of the sigmoid activation function to obtain the jump anomaly probability of the target energy meter at future times.

[0081] In one implementation, assuming the current time is time t, the formula for calculating the probability of the target energy meter's abnormal jump at a future time (time t+1) is as follows:

[0082]

[0083] in, represents the probability of an abnormal change in the target energy meter at a future time, b represents the preset bias value, and N represents the number of all times within the 10 minutes up to the current time. Since the sampling period in this embodiment of the invention is 1 minute, N=10. This represents the preset attenuation factor, and i represents the time interval, where one sampling period is one time interval. This is an exponential decay function with a preset decay factor as the base, and its value is the exponential decay weight. This represents the optimization coefficient for the jump anomaly at time ti. This represents the activation function.

[0084] It should be noted that, The larger the value, the greater the risk of abnormal fluctuations in the target electricity meter in the near future, and the stronger the persistence of these fluctuations. The larger the value, the higher the risk of the target energy meter experiencing abnormal fluctuations in the future. (Regarding the preset attenuation factor...) The settings, The larger the value, the slower the decay. The smaller the value, the faster the decay. Since this embodiment of the invention uses a recent short window for instantaneous rapid prediction—that is, prediction based on the optimized coefficient of abrupt changes within the last 10 minutes up to the current moment—the time window is relatively small, and the decay should not be too rapid. Therefore, the value is set... There are no restrictions here; implementers can set them according to the specific scenario.

[0085] Furthermore, based on the probability of abnormal fluctuations in the target energy meter at future times, anomaly warnings are issued for abnormal fluctuations in the target energy meter's power data at future times. Specifically:

[0086] The probability of a power meter not experiencing a jump anomaly within a preset time window prior to a future time is obtained and denoted as the historical jump anomaly probability. The percentile of the future jump anomaly probability among all historical jump anomaly probabilities is then calculated. If this percentile is greater than a preset percentile threshold, it indicates that the target power meter's health status at the current time significantly deviates from its historical normal pattern, posing a high risk of a jump anomaly. Therefore, an alarm is triggered at the current time to provide an anomaly warning for future power data jump anomalies in the target power meter. The preset time window ranges from 500 to 1000 seconds to ensure statistical significance, and the preset percentile threshold can be flexibly set between 85% and 95%. In this embodiment, the preset time window is set to 500 seconds and the preset percentile threshold to 85%, but this is not a limitation and can be set by the implementer according to the specific scenario.

[0087] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An energy meter with intelligent prediction function for abnormal power data fluctuations, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Based on the rated current, battery critical voltage, and maximum allowable clock deviation of the target energy meter, the A-phase current, battery voltage, and clock deviation between the built-in clock and the main clock of the target energy meter are piecewise linearly normalized at each moment within the preset time period up to the current moment to obtain the current abnormality score, voltage abnormality score, and clock deviation score of the target energy meter at each moment. Based on the physical mechanism of the fault, the coupling relationship between the voltage anomaly fraction and the clock deviation fraction at each moment and the current anomaly fraction at each moment are analyzed to obtain the jump anomaly coefficient of the target energy meter at each moment. According to the historical jump anomaly risk level of the target energy meter before each moment and the evolution rate of the jump anomaly coefficient at each moment, the jump anomaly coefficient at each moment is optimized to obtain the jump anomaly optimization coefficient of the target energy meter at each moment within the preset time period. Based on the anomaly optimization coefficient at each time point, the probability of the target energy meter experiencing anomalies at future time points is predicted, and the anomaly probability at future time points is obtained. Based on the anomaly probability, anomaly warning is issued for the power data anomalies of the target energy meter at future time points. The process of obtaining the jump anomaly coefficient of the target energy meter at each moment includes: For any given moment, calculate the product between the voltage anomaly fraction and the clock deviation fraction at that given moment to obtain the first anomaly level of the target energy meter at that given moment; Calculate the reciprocal of the sum of the current anomaly fraction at any given time and a preset constant, and use the negative of the product of the reciprocal and a preset sensitivity coefficient as the independent variable of the natural exponential function to obtain the second anomaly degree of the target energy meter at any given time. The first anomaly level and the second anomaly level are weighted and summed to obtain the jump anomaly coefficient of the target energy meter at any given time. The step of optimizing the jump anomaly coefficient at each moment based on the historical jump anomaly risk level of the target energy meter before each moment and the evolution rate of the jump anomaly coefficient at each moment, to obtain the optimized jump anomaly coefficient of the target energy meter at each moment within the preset time period, includes: Based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step, the jump anomaly risk momentum of the target energy meter at each time step is obtained. For any given moment, the product of the risk momentum of the jump anomaly at that moment and the preset sensitivity coefficient is used as the independent variable of the hyperbolic tangent function to obtain the optimization vector for optimizing the jump anomaly coefficient at that moment. The product between the preset optimization degree and the optimization vector is calculated, and the sum of the constant 1 and the product is used as the optimization factor for optimizing the jump anomaly coefficient at that moment. Calculate the product between the jump anomaly coefficient at any given time and the optimization factor to obtain the jump anomaly optimization coefficient of the target energy meter at any given time; The step of obtaining the jump anomaly risk momentum of the target energy meter at each time step based on the historical jump anomaly risk level of the target energy meter before each time step and the evolution rate of the jump anomaly coefficient at each time step includes: For any given moment, based on the jump anomaly coefficient of the target energy meter at each moment from the start of operation to that moment, the historical jump anomaly risk quality of the target energy meter up to that moment is obtained using an exponential weighted average algorithm; Subtracting the jump anomaly coefficient of the previous moment from the jump anomaly coefficient at any given moment yields the instantaneous change rate of the target energy meter at any given moment. Calculate the product between the historical jump anomaly risk quality and the instantaneous change rate of the jump anomaly to obtain the risk characteristic value of the target energy meter at any given moment; During the historical normal operation of the target energy meter, the risk quality of the jump anomaly and the instantaneous change rate of the jump anomaly at each historical normal moment within the historical normal operation period of a preset duration are obtained. The target energy meter does not experience any power data jump anomaly during the historical normal operation period, and the A-phase current, battery voltage and clock deviation of the target energy meter at each historical normal moment do not exceed the rated current, battery critical voltage and maximum allowable clock deviation of the target energy meter. Calculate the standard deviation of the risk quality of the jump anomaly at all historical normal times, and denot it as the risk quality standard deviation. Calculate the standard deviation of the instantaneous change rate of the jump anomaly at all historical normal times, and denot it as the instantaneous change rate standard deviation. Calculate the product between the risk quality standard deviation and the instantaneous change rate standard deviation. Use a preset constant and the product as the denominator, and use the risk characteristic value of the target energy meter at any time as the numerator to obtain the jump anomaly risk momentum of the target energy meter at any time. The jump anomaly risk momentum is a vector. The step of predicting the probability of a jump anomaly in the target energy meter at a future time based on the jump anomaly optimization coefficient at each time moment, and obtaining the jump anomaly probability at the future time moment, includes: Based on the time interval between each moment of the target energy meter within the preset time period and the current moment, the exponential decay weight of each moment within the preset time period is obtained using an exponential decay function with a preset decay factor as the base. Based on the exponential decay weight of each moment within the preset time period, the jump anomaly optimization coefficients of the target energy meter at each moment within the preset time period are weighted and summed to obtain the summation result. The summation result and the preset bias value are used as the independent variable of the sigmoid activation function to obtain the jump anomaly probability of the target energy meter at future moments.

2. The energy meter with intelligent prediction function for abnormal power data fluctuations according to claim 1, characterized in that, The method of providing early warning of abnormal power data fluctuations in the target electricity meter at future times based on the abnormal fluctuation probability includes: The probability of a jump anomaly in the target energy meter at a historical moment when no jump anomaly occurred within a preset time window before the future moment is obtained and recorded as the historical jump anomaly probability. The percentile of the jump anomaly probability at the future moment is obtained among all historical jump anomaly probabilities. If the percentile is greater than a preset percentile threshold, an anomaly warning is issued for the power data jump anomaly of the target energy meter at the future moment.

Images