A method for monitoring operation data of a 400V frame circuit breaker

Abstract

This invention relates to the field of frame circuit breaker operation monitoring technology, specifically to a method for monitoring the operation data of a 400V frame circuit breaker. The method includes: collecting different types of operation data and obtaining the observation vector and state vector of the EKF (Extended Kernel Factor), thus obtaining the forward neighbor dataset for each type of operation data at the current time; obtaining the confidence index and dependency bias factor for the current time based on the forward neighbor dataset for each type of operation data; obtaining the chaotic change characteristic value for the current time based on the prediction covariance matrix of each time step in the forward neighbor dataset; obtaining the proportion of historical outliers at the current time based on the outlier data in the forward neighbor dataset; correcting the process noise covariance matrix for the current time step to obtain the process noise covariance matrix for the next time step, and then obtaining the state vector for the next time step, thereby monitoring the operation status of the frame circuit breaker. This application can effectively monitor the operation status of frame circuit breakers.

Description

Technical Field

[0001] This invention relates to the field of operation monitoring technology for frame circuit breakers, and specifically to a method for monitoring the operation data of a 400V frame circuit breaker. Background Technology

[0002] A 400V frame circuit breaker is a low-voltage electrical appliance used for circuit protection, featuring overload protection, short-circuit protection, undervoltage protection, and leakage current protection. It is suitable for three-phase AC systems with a rated operating voltage of AC 400V and is commonly found in industrial and residential power distribution systems.

[0003] Frame circuit breakers are key protection devices in low-voltage power distribution systems, responsible for isolating faults such as short circuits and overloads. Therefore, it is necessary to monitor their operating data to prevent circuit breakers from refusing to operate or operating erroneously during faults, thereby providing system safety, ensuring power supply reliability, reducing operation and maintenance costs, and supporting the construction of smart grids.

[0004] Since critical states of frame circuit breakers during operation (such as contact temperature and contact resistance) cannot be directly measured by sensors, as they are mostly located inside the circuit breaker, installing sensors would damage the circuit breaker's insulation structure or openings, interfering with normal power supply. Wireless sensors have poor reliability in high-current electromagnetic environments and are very expensive. Therefore, extended Kalman filters (EKF) are typically used to infer these unobservable critical states from observable states, enabling safer and more reliable monitoring of frame circuit breakers.

[0005] EKF (Electronic Kinematics Flow) requires balancing model predictions with actual observations to obtain the final state estimate. In this process, the size of the process noise covariance matrix Q determines whether the final output depends more on model predictions or actual observations. However, since the frame circuit breaker itself is subject to factors such as load fluctuations, voltage surges, and changes in ambient temperature, these unobservable states are constantly changing with time and operating conditions. The impact of model predictions and actual observations on the final output is also constantly changing. In the traditional approach, the size of the elements in Q in EKF is always kept constant. This means that under different environments and conditions, the impact of model predictions and actual observations on the final output is always kept constant, which cannot adapt to these sudden disturbances and may lead to the state estimate deviating from the actual state. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention aims to provide a method for monitoring the operational data of a 400V frame circuit breaker. The specific technical solution adopted is as follows:

[0007] One embodiment of the present invention provides a method for monitoring the operating data of a 400V frame circuit breaker, the method comprising:

[0008] Collect different types of operational data from the frame circuit breaker and obtain the observation vector and state vector of the EKF; obtain the forward neighbor dataset for each type of operational data at the current time.

[0009] The confidence index for the current moment is obtained based on the variation and fluctuation of the residuals of various running data at each time step in the forward neighbor dataset of each type of running data at the current moment; the dependency bias factor for the current moment is determined based on the confidence index.

[0010] Obtain the chaotic change feature value at the current time based on the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset; obtain the historical outlier ratio at the current time based on the number of outliers in each type of running data in the forward neighbor dataset for each type of running data;

[0011] The process noise covariance matrix of the EKF at the current time is corrected based on the dependency bias factor, chaotic change eigenvalues ​​and the proportion of historical outliers at the current time to obtain the process noise covariance matrix at the next time.

[0012] The state vector at the next moment is obtained using EKF based on the process noise covariance matrix and observation vector at the next moment; the operating status of the frame circuit breaker is monitored based on the state vector at the next moment.

[0013] Preferably, the observation vector and state vector of the EKF are obtained, including:

[0014] The different types of operational data at a given moment are combined to form the observation vector of the EKF at that moment; the contact temperature and contact resistance at each moment are combined to form the state vector of the EKF at each moment.

[0015] Preferably, the forward neighbor dataset for each type of running data at the current time is obtained, including:

[0016] Starting with the current running data of a certain type of running data, the system sequentially collects running data within a preset time period to form the forward neighbor dataset of that type of running data at the current moment.

[0017] Preferably, the confidence index for the current moment is obtained based on the variation and fluctuation of the residuals of various running data at each time step in the forward neighbor dataset of each type of running data at the current moment, including:

[0018] A first mapping value is obtained by negatively correlating the absolute value of the difference between the residual of a certain type of running data at the current time and the median of the residuals of the same type of running data at each time step in the forward neighbor dataset using an exponential function with a base of the natural constant. The time interval from a time step in the forward neighbor dataset to the current time step is added to the first preset value, and the reciprocal is calculated to obtain the reciprocal of the time interval at that time step. The reciprocal of the time interval at each time step in the forward neighbor dataset is compared with the sum of the reciprocals of the time intervals at all times to obtain the weight of each time step. The weights of each time step are then used to calculate the weights of the residuals at each time step in the forward neighbor dataset for the current type of running data. The weighted average of the residuals of the operational data of this type is used to calculate the weighted variance of the forward neighbor dataset of the operational data of this type at the current time, based on the weighted average and the weight of each time step in the forward neighbor dataset. The weighted variance of the forward neighbor dataset of the operational data of this type at the current time is negatively correlated using an exponential function with the natural constant as the base to obtain a second mapping value. The average of the first mapping value and the second mapping value of the forward neighbor dataset of the operational data of this type at the current time is calculated to obtain the residual change index of the operational data of this type at the current time. The average of the residual change indices of all types of operational data at the current time is used to obtain the confidence index of the current time.

[0019] Preferably, determining the dependency bias factor at the current moment based on the confidence index includes:

[0020] If the confidence index at the current moment is less than or equal to the reference threshold, the dependency bias factor is set to 1; if the confidence index at the current moment is greater than the reference threshold, the dependency bias factor is set to -1.

[0021] Preferably, the chaotic change feature value at the current time is obtained based on the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset, including:

[0022] The diagonal sum of the predicted covariance matrix at the current time is obtained by summing the diagonal elements at the current time. The third mapping value is obtained by mapping the diagonal sum of the current time using the hyperbolic tangent function. The absolute value of the difference between the diagonal sums of two adjacent times in the forward neighbor dataset is divided by the diagonal sum of the next time after those two adjacent times to obtain the rate of change of the diagonal sum. The fourth mapping value is obtained by negatively mapping the sum of the rates of change of all diagonal sums in the forward neighbor dataset using an exponential function with the natural constant as the base. The average of the third and fourth mapping values ​​is used to obtain the chaotic change feature value at the current time.

[0023] Preferably, the proportion of historical outliers at the current moment is obtained based on the number of outliers in the forward neighbor dataset of each type of running data, including:

[0024] The historical outlier ratio at the current moment is obtained by comparing the sum of the number of outliers in each type of running data in the forward neighbor dataset with the total number of all running data in the forward neighbor dataset of all types of running data.

[0025] Preferably, the process noise covariance matrix of the EKF at the current time is corrected based on the dependency bias factor, disordered change eigenvalues, and the proportion of historical outliers at the current time to obtain the process noise covariance matrix at the next time, including:

[0026] The adjustment magnitude of the process noise covariance matrix at the next moment is obtained by taking the average of the chaotic change characteristic value and the proportion of historical outliers at the current moment; the correction coefficient is obtained by multiplying the dependence bias factor at the current moment and the adjustment magnitude of the process noise covariance matrix at the next moment and adding it to the first preset value; and the process noise covariance matrix at the next moment is obtained by multiplying each element of the process noise covariance matrix at the current moment with the correction coefficient.

[0027] The embodiments of this invention have at least the following beneficial effects: This application collects different types of operating data from a frame circuit breaker, thereby determining the observation vector and state vector of the EKF, and obtaining the forward neighbor dataset for each type of operating data at the current time; then, it obtains the confidence index for the current time by analyzing the variation and fluctuation of the residuals of various operating data at each time in the forward neighbor dataset for each type of operating data, thereby determining the dependency bias factor for the current time and guiding the adjustment direction of the process noise covariance matrix; next, it analyzes the chaotic change characteristic value of the current time by analyzing the diagonal elements in the prediction covariance matrix at each time in the forward neighbor dataset, and obtains the proportion of historical outliers at the current time based on the number of abnormal data in each type of operating data in the forward neighbor dataset for each type of operating data; then, it utilizes... The process noise covariance matrix of the EKF at the current moment is corrected using the dependency bias factor, chaotic change eigenvalues, and the proportion of historical outliers to obtain the process noise covariance matrix at the next moment. Finally, based on the corrected process noise covariance matrix and combined with the EKF, the state vector at the next moment is obtained, thereby monitoring the operating status of the frame circuit breaker. By analyzing the internal change characteristics of the EKF, the size of the elements in the process noise covariance matrix is ​​adjusted in real time, enabling the EKF to respond promptly to sudden disturbances or noises (such as current surges or sudden changes in ambient temperature) that may occur during the operation of the frame circuit breaker. The state estimation is closer to the actual operating state, ensuring that the monitoring system can provide reliable and continuous key state information under various operating conditions, and achieving effective monitoring of operating data. Attached Figure Description

[0028] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating a method for monitoring the operational data of a 400V frame circuit breaker, as provided in an embodiment of the present invention. Detailed Implementation

[0030] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for monitoring the operation data of a 400V frame circuit breaker according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0032] The following describes in detail, with reference to the accompanying drawings, a specific scheme for monitoring the operation data of a 400V frame circuit breaker provided by the present invention.

[0033] Example: The main application scenario of this invention is: This application mainly monitors the operating data of frame circuit breakers.

[0034] Please see Figure 1 The document illustrates a flowchart of a method for monitoring the operational data of a 400V frame circuit breaker according to an embodiment of the present invention. The method includes the following steps:

[0035] Step S1: Collect different types of operating data of the frame circuit breaker and obtain the observation vector and state vector of EKF; obtain the forward neighbor dataset of each type of operating data at the current time.

[0036] This application requires collecting data that can be directly obtained during the operation of the frame circuit breaker, using it as operational data. Since there are multiple types of operational data, it is necessary to collect different types of operational data at each moment, such as current, voltage, and casing temperature during the operation of the frame circuit breaker. During data collection, the acquisition frequency needs to be synchronized to adapt to EKF requirements, and timestamp synchronization is required to ensure that the above-mentioned observation data have a unified time base. Thus, the different types of operational data at a given moment are combined to form the EKF observation vector for that moment.

[0037] It should be noted that the operational data that makes up the observation vector all need to be preprocessed in advance, mainly including: denoising, removing extreme outliers using thresholding or statistical outlier detection, and replacing them with values ​​collected at the previous time step. Missing data is handled using methods such as linear interpolation.

[0038] Furthermore, it is necessary to determine the key states, which are crucial to the health of the frame circuit breaker but cannot be directly measured. These are set according to the scenario and business requirements. This application provides the most critical candidate states in the current scenario: contact temperature and contact resistance. That is, the contact temperature and contact resistance at each moment constitute the state vector of the EKF at each moment.

[0039] Since the changes in contact temperature and resistance in a frame circuit breaker tend to accumulate gradually, taking several minutes to show a clear trend, this application requires obtaining operational data up to the current moment for analysis. Specifically, starting with the current operational data for a particular type of data, operational data within a preset time period is sequentially collected to form a forward neighbor dataset for that type of operational data at the current moment. The preset time period is typically 5 minutes and can be adjusted according to actual conditions. This allows for the acquisition of forward neighbor datasets for each type of operational data at the current moment, which then enables subsequent analysis. The forward neighbor dataset includes the operational data at the current moment.

[0040] Step S2: Obtain the confidence index for the current time based on the changes and fluctuations of the residuals of various running data at each time in the forward neighbor dataset of each type of running data at the current time; determine the dependency bias factor for the current time based on the confidence index.

[0041] Since increasing or decreasing the "process noise covariance matrix Q" will have completely opposite effects, it is necessary to first clarify the direction of adjustment to ensure that the entire adjustment behavior is correct and effective.

[0042] In the standard EKF, the estimation of the state vector at each step is obtained by combining model prediction and observation correction. The model prediction depends on the system's state equation and the state at the previous time step, while the observation correction depends on the sensor's measurement data. The final EKF estimation result (the estimation of the state vector) is affected by the interaction of the two aspects.

[0043] According to the EKF state update formula, the observation correction is obtained by combining the Kalman gain and the residual. The Kalman gain is mainly affected by the "process noise covariance matrix Q". The Kalman gain determines the weight of the residual in the state update. According to their influence relationship, the adjustment direction of the process noise covariance matrix Q is actually to determine the degree of influence of the observation correction on the final estimation result.

[0044] The purpose of observation correction is to use the difference between the sensor's observed data and the prediction (i.e., the residual) and combine it with the Kalman gain to correct the prediction result and obtain a more accurate state estimate. Therefore, the variation characteristics of the residual can serve as a direct basis for adjusting Q. During circuit breaker operation, if the residual is generally small and fluctuates stably, it indicates that the model estimation can reflect the actual state well. In this case, Q can be reduced to make the estimation result rely more on the model prediction, thereby reducing the sensitivity to observation noise. If the residual is large and fluctuates drastically, it indicates that there is a significant deviation between the model prediction and the actual observation, such as when the circuit breaker is under high load, the contacts are overheating, or the contact resistance changes abruptly. In this case, the reliability of the model prediction decreases, and Q needs to be increased to make the role of observation correction in the final estimate greater, so as to ensure that the state estimate does not deviate from the true situation in the long term.

[0045] In the EKF state update formula (as shown below, this is the existing formula):

[0046] ,

[0047] ,

[0048] in, This represents the updated state vector estimate. This represents a prediction of the state, but only uses information from time k-1. The Kalman gain ("process noise covariance matrix Q") is expressed as the sum of the Kalman gain and the Kalman noise covariance matrix Q. (positively correlated) This represents the observation residual, which is the residual of a type of running data at time k. This represents the actual observations (operational data) acquired by the sensors, specifically the operational data at time k. This represents the predicted observation, which is a type of operational data predicted by EKF at time k.

[0049] Based on the above idea, we analyze the residual fluctuations of various operating data at each time step in the forward neighbor dataset of each type of operating data at the current time to obtain the confidence index of the residual part at the current time.

[0050] Specifically, a first mapping value is obtained by negatively correlating the absolute value of the difference between the residual of a certain type of operational data at the current time and the median of the residuals of the same type of operational data at each time step in the forward neighbor dataset using an exponential function with a base of the natural constant. The time interval between a time step in the forward neighbor dataset and the current time step is added to a first preset value, and the reciprocal of this time interval is obtained. The reciprocal of the time interval at each time step in the forward neighbor dataset is then compared with the sum of the reciprocals of the time intervals at all times to obtain the weight of each time step. Finally, the weights of each time step are used to calculate the weights of the residuals of the same type of operational data at the current time step in the forward neighbor dataset. The weighted average of the residuals of the operational data at time step 1 is used to calculate the weighted variance of the forward neighbor dataset of the operational data at the current time step 2 based on the weighted average and the weight of each time step in the forward neighbor dataset. The weighted variance of the forward neighbor dataset of the operational data at the current time step 2 is negatively correlated and mapped using an exponential function with the natural constant as the base to obtain a second mapping value. The average of the first mapping value and the second mapping value of the forward neighbor dataset of the operational data at the current time step 3 is calculated to obtain the residual change index of the operational data at the current time step 4. The average of the residual change indices of all types of operational data at the current time step 4 is calculated to obtain the confidence index of the current time step 5.

[0051] The specific calculation model for the confidence index is as follows:

[0052] ,

[0053] in, represents the confidence index at the current moment; M represents the number of data types; exp represents the exponential function with the natural constant as the base, used for negative correlation mapping, and exp(-) is the inverse proportional normalization function, which ensures that the larger the final output result, the smaller the current residual and the more stable the fluctuation, and the higher the confidence; conversely, the smaller the current residual and the more drastic the fluctuation, the lower the confidence. This represents the residual of the α-th type of running data at the k-th time (current time). This represents the median of the residuals of the α-th type of running data at each time step of the forward neighbor dataset of the α-th type of running data at time step k (the current time). This represents the first mapped value. Since the residuals may fluctuate significantly, the median is used here to prevent the fluctuations from affecting the relative size of the residuals. The smaller the absolute value of the difference, the closer the residual is to the overall numerical characteristics at the current moment, and the smaller the residual is. This represents the weighted variance of the forward neighbor datasets of the α-th type of running data at time k (the current time). Indicates the second mapping value; ∆ represents the residual change index of the α-th type of running data at the k-th time (current time); ∆ represents the number of data in a forward neighbor dataset, or the number of time points in the set.

[0054] Weighted variance The specific calculation model is as follows:

[0055] ,

[0056] Where k represents the k-th moment, which is the current moment; This represents the reciprocal of the time interval at the i-th moment in the forward neighbor dataset. The reason for adding the denominator to the first preset value is to prevent the denominator from being 0. ki represents the time interval between the i-th moment and the k-th moment in the forward neighbor dataset. The first preset value is 1. This represents the weight of the i-th time in the forward neighbor dataset. The smaller the time distance between a time and the k-th time in the forward neighbor dataset, the greater the weight of that time. In the current scenario, since the critical state of the circuit breaker may change rapidly with load and operation, the reference value of historical residuals to the current state is different: the residuals closer to the current time can better reflect the true state of the system, while the information of earlier residuals is lagging. Therefore, a time decay weight is introduced in the variance calculation, so that the weight of recent residuals is larger and the weight of historical residuals is smaller, which more accurately describes the volatility of the current residuals. Let represent the residual of the i-th time-n in the forward neighbor dataset of the k-th time-n (current time) type of running data. It represents the weighted average of the residuals of a type of running data at each time step in the forward neighboring datasets of the current running data. The smaller the weighted variance, the smaller the recent residual fluctuation, indicating that the model prediction is closer to and more stable than the observed value. Conversely, the larger the recent residual fluctuation, the larger the recent residual fluctuation, indicating that the model prediction deviates significantly from the observed value and is unstable.

[0057] The confidence index represents the confidence level of the residual at the current time, reflecting the relationship between the current time and the previous time. The variation characteristics of the time residuals; further, based on the confidence index, the EKF dependence bias factor with respect to observation correction is obtained, when When the value of Q is large, it is assumed that the current residual value is small and the fluctuation is stable. Therefore, Q can be reduced to allow the estimation results to rely more on model predictions, thereby reducing sensitivity to observation noise and decreasing the dependence on the bias factor. A value of -1 indicates a large residual with significant fluctuations, reducing the reliability of the model's predictions. In such cases, Q needs to be increased to maximize the role of observation corrections in the final estimate, thus increasing the dependence on the bias factor. The threshold is 1. The specific threshold should be set according to actual needs, as... The value range is [0,1]. Therefore, this scheme takes a compromise and sets the reference threshold to 0.5. If the confidence index at the current time is less than or equal to the reference threshold, the dependency bias factor is set to 1. If the confidence index at the current time is greater than the reference threshold, the dependency bias factor is set to -1.

[0058] This determines the direction of change of elements in the process noise covariance matrix Q at the current moment.

[0059] Step S3: Obtain the chaotic change feature value at the current moment based on the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset; obtain the historical outlier ratio at the current moment based on the number of outliers in each type of running data in the forward neighbor dataset for each type of running data.

[0060] The previous step obtained the EKF dependency bias factor for observation correction. Based on the dependency bias factor, the adjustment direction of the "process noise covariance matrix Q" can be determined. Therefore, in this step, the adjustment magnitude of Q is considered through feature analysis.

[0061] The estimation of the state vector is the core object of EKF output. The state vector is essentially a state that cannot be directly measured and can only be inferred by EKF. Since it is an inference, there must be uncertainty. EKF uses the prediction covariance matrix P to express the uncertainty of the estimation. The magnitude of Q determines how much initial "uncertainty" is given to the EKF model in the prediction stage. That is, Q directly affects P, and P expresses the uncertainty of the estimated state. Therefore, the change characteristics of uncertainty can be analyzed as a basis for adjusting the magnitude of Q.

[0062] Where P is a symmetric positive definite matrix containing the variance (this variance is not the variance calculated in data statistics, but is specifically calculated in EKF, though it is also called variance) and covariance of all state components. Its diagonal elements, i.e. the variance of each state component, directly represent the uncertainty of the estimation of each state component. Its changing trend directly reflects the change in EKF's "confidence" in the state estimation. When its change is stable, it means that the estimation is relatively stable and this stable estimation accuracy should be maintained. Therefore, only a small adjustment is needed to Q. Conversely, when the change is chaotic, it means that the system estimation is unstable and the state estimation is insufficient to explain the current circuit breaker state. Q needs to be increased more to feed back this uncertainty to EKF.

[0063] Based on the above idea, the chaotic change characteristic value of the state component variance is obtained. The chaotic change characteristic value at the current time is obtained from the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset.

[0064] Specifically, the diagonal elements of the current time step are summed to obtain the diagonal element sum at the current time step; the hyperbolic tangent function is used to map the diagonal element sum at the current time step to obtain the third mapping value; the absolute value of the difference between the diagonal element sums of two adjacent times in the forward neighbor dataset is divided by the diagonal element sum of the next time step of those two adjacent times to obtain the rate of change of the diagonal element sum; the sum of the rates of change of all diagonal element sums in the forward neighbor dataset is negatively correlated using an exponential function with the natural constant as the base to obtain the fourth mapping value; the average of the third mapping value and the fourth mapping value is calculated to obtain the chaotic change feature value at the current time step.

[0065] The calculation model for the chaotic change characteristic value at the current moment is as follows:

[0066] ,

[0067] Let Q represent the chaotic variation characteristic value of the variance of the state component at the current time (the k-th time). tanh() represents the hyperbolic tangent function, and exp()- represents the inverse proportional normalization function. The larger the final output result, the larger the variance of the current state component. At the same time, the chaotic variation indicates that the system estimation is unstable and the state estimation is insufficient to explain the current circuit breaker state. Q needs to be increased more to feed back this uncertainty to EKF. Conversely, if the estimation is stable, it indicates that the estimation is stable and this stable estimation accuracy should be maintained. Therefore, only a small adjustment is needed to Q.

[0068] in, Let be the diagonal element in the i-th row and i-th column of the prediction covariance matrix at the current time (k-th time), representing the variance of the state component. The sum of these elements reflects the overall uncertainty of the current state component estimation; the larger the sum, the higher the uncertainty. This is the sum of the diagonal elements in the prediction covariance matrix at the current time, which is also the sum of the diagonal elements at the current time. Indicates the third mapping value; Indicates the first in the forward neighbor dataset The difference between the overall variance (sum of diagonal elements) of the state vector at time step one and the previous time step. The sum of these sums represents the rate of change of the state vector variance over the time interval (k, ∆). The larger the sum, the more chaotic the variance change of the state vector over this time interval.

[0069] Because the operating data of the frame circuit breaker itself is affected by sudden load changes, equipment abnormalities or measurement noise, outliers will appear in the observation vector. Regardless of whether these outliers belong to noise or actual operating condition abnormalities, they will affect the final estimated state output of the EKF. The higher the proportion of outliers, the less accurately the EKF can estimate the state vector, and the more significant the adjustments are needed to enable the EKF to react quickly to such sudden changes.

[0070] Meanwhile, the chaotic variation characteristic value of the variance of the state component reflects the continuous change of uncertainty over time. However, there are occasional anomalies that interfere with the EKF, and the variance variation may not be able to fully capture such disturbances. Therefore, the number of anomalous data in each running data in the forward neighbor dataset of each running data is obtained. The anomalous data in each running data in the forward neighbor dataset of each running data can be directly determined according to the 3σ principle.

[0071] The proportion of historical outliers at the current moment is obtained by comparing the sum of the number of outliers in the forward neighbor datasets for each type of running data with the total number of outliers in the forward neighbor datasets for all types of running data. This proportion is denoted as [missing information]. A higher proportion of historical outliers indicates a greater percentage of outliers in the historical data during that period.

[0072] Step S4: Based on the dependency bias factor, chaotic change eigenvalues ​​and the proportion of historical outliers at the current time, the process noise covariance matrix of EKF at the current time is corrected to obtain the process noise covariance matrix at the next time.

[0073] The above obtains the dependency bias factor, disordered change eigenvalues, and historical outlier ratios at the current moment. The dependency bias factor determines the adjustment direction of the element sizes in the process noise covariance matrix. Furthermore, the adjustment magnitude is determined based on the disordered change eigenvalues ​​and the historical outlier ratios.

[0074] Specifically, the adjustment magnitude of the process noise covariance matrix at the next time step is obtained by taking the mean of the current chaotic change eigenvalues ​​and the proportion of historical outliers, denoted as . .

[0075] Then, the process noise covariance matrix at the current time is corrected based on the dependency bias factor at the current time and the adjustment magnitude of the process noise covariance matrix at the next time.

[0076] Specifically, the current time-dependent bias factor and the adjustment magnitude of the process noise covariance matrix at the next time-instance are multiplied together and added to a first preset value to obtain a correction coefficient; each element in the process noise covariance matrix at the current time-instance is multiplied by the correction coefficient to obtain the process noise covariance matrix at the next time-instance.

[0077] The specific calculation model for the process noise covariance matrix at the next time step is as follows:

[0078] ,

[0079] in, This represents the process noise covariance matrix at the next time step. This represents the process noise covariance matrix at the current moment. This represents the correction factor. This indicates the adjustment magnitude of the process noise covariance matrix at the next time step. This represents the dependency bias factor at the current moment, with a first preset value of 1.

[0080] Step S5: Based on the process noise covariance matrix and observation vector at the next time step, the state vector at the next time step is obtained using EKF; the operating status of the frame circuit breaker is monitored according to the state vector at the next time step.

[0081] The process noise covariance matrix for the next time step is obtained above. Furthermore, the state vector for the next time step can be obtained by using EKF based on the observation vector at the next time step.

[0082] Finally, the operating status of the frame circuit breaker is monitored based on the state vector at the next moment. Specifically, according to the circuit breaker equipment design specifications or safety thresholds, each key state component (each element, contact temperature and contact resistance) in the state vector is judged, the estimated values ​​of multiple state components are statistically analyzed, the health or risk level of the current operating status is assessed, and the monitoring results are used for real-time alarms, operation and maintenance decisions or preventive maintenance.

[0083] Through the above steps, the process noise covariance matrix at the next moment is adaptively adjusted. Based on the weights of prediction and observation correction in the dynamic equilibrium model of actual system uncertainty, EKF can respond in a timely manner to sudden disturbances or noises (such as current surges or sudden changes in ambient temperature) that may occur during the operation of the frame circuit breaker. The state estimation is closer to the actual operating state, ensuring that the monitoring system can provide reliable and continuous key state information under various operating conditions, and realizing effective monitoring of operating data.

[0084] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0085] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the operational data of a 400V frame circuit breaker, characterized in that, The method includes: Collect different types of operational data from the frame circuit breaker and obtain the observation vector and state vector of the EKF; obtain the forward neighbor dataset for each type of operational data at the current time. The confidence index for the current moment is obtained based on the variation and fluctuation of the residuals of various running data at each time step in the forward neighbor dataset of each type of running data at the current moment; the dependency bias factor for the current moment is determined based on the confidence index. Obtain the chaotic change feature value at the current moment based on the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset; obtain the historical outlier ratio at the current moment based on the number of outliers in each type of running data in the forward neighbor dataset for each type of running data; The process noise covariance matrix of the EKF at the current time is corrected based on the dependency bias factor, chaotic change eigenvalues ​​and the proportion of historical outliers at the current time to obtain the process noise covariance matrix at the next time. The state vector at the next time step is obtained using EKF based on the process noise covariance matrix and observation vector at the next time step; the operating status of the frame circuit breaker is monitored based on the state vector at the next time step. The step of obtaining the confidence index at the current moment based on the variation and fluctuation of the residuals of various running data at each time step in the forward neighbor dataset of each type of running data at the current moment includes: A first mapping value is obtained by negatively correlating the absolute value of the difference between the residual of a certain type of running data at the current time and the median of the residuals of the same type of running data at each time step in the forward neighbor dataset using an exponential function with a base of the natural constant. The time interval from a time step in the forward neighbor dataset to the current time step is added to the first preset value, and the reciprocal is calculated to obtain the reciprocal of the time interval at that time step. The reciprocal of the time interval at each time step in the forward neighbor dataset is compared with the sum of the reciprocals of the time intervals at all times to obtain the weight of each time step. The weights of each time step are then used to calculate the weights of the residuals at each time step in the forward neighbor dataset for the current type of running data. The weighted average of the residuals of the operational data of this type is used to calculate the weighted variance of the forward neighbor dataset of the operational data of this type at the current time, based on the weighted average and the weight of each time step in the forward neighbor dataset. A second mapping value is obtained by negatively correlated mapping of the weighted variance of the forward neighbor dataset of the operational data of this type at the current time using an exponential function with the natural constant as the base. The average of the first and second mapping values ​​of the forward neighbor dataset of the operational data of this type at the current time is used to obtain the residual change index of the operational data of this type at the current time. The confidence index of the current time is obtained by averaging the residual change indices of all types of operational data at the current time. The step of obtaining the chaotic change feature value at the current time based on the diagonal elements of the prediction covariance matrix at each time step in the forward neighbor dataset includes: The diagonal sum of the predicted covariance matrix at the current time is obtained by summing the diagonal elements at the current time. The third mapping value is obtained by mapping the diagonal sum of the current time using the hyperbolic tangent function. The absolute value of the difference between the diagonal sums of two adjacent times in the forward neighbor dataset is divided by the diagonal sum of the next time after those two adjacent times to obtain the rate of change of the diagonal sum. The fourth mapping value is obtained by negatively mapping the sum of the rates of change of all diagonal sums in the forward neighbor dataset using an exponential function with the natural constant as the base. The average of the third and fourth mapping values ​​is used to obtain the chaotic change feature value at the current time.

2. The method for monitoring the operation data of a 400V frame circuit breaker according to claim 1, characterized in that, The obtained observation vector and state vector of EKF include: The different types of operational data at a given moment are combined to form the observation vector of the EKF at that moment; the contact temperature and contact resistance at each moment are combined to form the state vector of the EKF at each moment.

3. The method for monitoring the operation data of a 400V frame circuit breaker according to claim 1, characterized in that, The process of obtaining the forward neighbor dataset for each type of running data at the current moment includes: Starting with the current running data of a certain type of running data, the system sequentially collects running data within a preset time period to form the forward neighbor dataset of that type of running data at the current moment.

4. The method for monitoring the operation data of a 400V frame circuit breaker according to claim 1, characterized in that, The determination of the dependency bias factor at the current moment based on the confidence index includes: If the confidence index at the current moment is less than or equal to the reference threshold, the dependency bias factor is set to 1; if the confidence index at the current moment is greater than the reference threshold, the dependency bias factor is set to -1.

5. The method for monitoring the operation data of a 400V frame circuit breaker according to claim 1, characterized in that, The method of obtaining the proportion of historical outliers at the current moment based on the number of outliers in each type of running data in the forward neighbor dataset for each type of running data includes: The historical outlier ratio at the current moment is obtained by comparing the sum of the number of outliers in each type of running data in the forward neighbor dataset with the total number of all running data in the forward neighbor dataset of all types of running data.

6. The method for monitoring the operation data of a 400V frame circuit breaker according to claim 1, characterized in that, The process noise covariance matrix of the EKF at the current time moment is corrected based on the dependency bias factor, chaotic change eigenvalues, and the proportion of historical outliers at the current time moment to obtain the process noise covariance matrix at the next time moment, including: The adjustment magnitude of the process noise covariance matrix at the next moment is obtained by taking the average of the chaotic change characteristic value and the proportion of historical outliers at the current moment; the correction coefficient is obtained by multiplying the dependence bias factor at the current moment and the adjustment magnitude of the process noise covariance matrix at the next moment and adding it to the first preset value; and the process noise covariance matrix at the next moment is obtained by multiplying each element of the process noise covariance matrix at the current moment with the correction coefficient.

Images