Method for monitoring operation state of fuse
By extracting and correcting abnormal monitoring values of fuses using the isolated forest algorithm, the accuracy problem of fuse operating parameter monitoring in complex environments is solved, and higher reliability of status judgment is achieved.
Patent Information
- Application Number
- CN202511875874.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-12
- Publication Date
- 2026-01-20
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In complex environments, abnormal values can easily be generated in the monitoring of fuse operating parameters, affecting technicians' accurate judgment of the operating status.
An isolated forest algorithm is used to extract abnormal monitoring values from the parameter sequence and correct these values, including fixing offsets, mutations, and missing monitoring values. The corrected parameter sequence is then used to determine the operating status of the fuse.
It improves the accuracy and reliability of fuse operation status judgment and reduces the interference of abnormal values on the judgment.
Smart Images
Figure FT_1 
Figure SMS_1 
Figure SMS_4
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the field of intelligent power distribution system, in particular, relates to a method for monitoring the running state of a fuse. BACKGROUND
[0002] The fuse is a device for circuit protection, the main function is to automatically cut off the current when the circuit is overloaded or short-circuited, so as to prevent the damage of electrical equipment or the occurrence of fire accidents, etc., because it usually also integrates the modules for monitoring, controlling, communicating, etc. of the electrical system, it is also often used in intelligent power distribution system.
[0003] During the operation of the fuse, the related parameters in the operation process need to be monitored and recorded to obtain the monitoring values, which can be analyzed by the operating personnel to analyze the running state change process of the fuse, but when the fuse faces a complex environment, for example, when the fuse is installed and used outdoors, the monitored and recorded running parameters of the fuse are prone to abnormal values, which in turn affects the accuracy of the subsequent technical personnel in judging the running state of the fuse under the running parameter. Therefore, a monitoring method is needed to reduce the interference of abnormal values to technical personnel, and to improve the accuracy of technical personnel in judging the running state of the fuse under a certain running parameter. SUMMARY
[0004] The existing monitoring method for the running state of the fuse is prone to abnormal values during the monitoring of the running parameters, which reduces the accuracy of the technical personnel in judging the running state of the running parameters. To solve this problem, the following invention is proposed: A method for monitoring the running state of a fuse, comprising the following steps: Step S1, determining the running parameters of the fuse, the value obtained by monitoring the running parameters is called monitoring value, and the monitoring values are arranged in time sequence as parameter sequence; Step S2, dividing the monitoring values in the parameter sequence into normal monitoring values and abnormal monitoring values; Step S3, correcting the abnormal monitoring values and obtaining the corrected parameter sequence; Step S4, classifying the running state of the running parameters, and using the corrected parameter sequence to judge the current running state of the running parameters.
[0005] Further, in step S1, the running parameters are periodic parameters or non-periodic parameters, and the parameter sequence of the non-periodic parameters and the non-periodic parameters is recorded as X, which is represented as follows,
[0006] Wherein, if the running parameters are non-periodic parameters, x iXi represents the monitoring value of the monitoring parameter at the ith time point; if the monitoring parameter is a periodic parameter, the parameter sequence X is composed of the monitoring values at the same time point in different periods in the order of the periods, and xi represents the monitoring value of the ith period at a certain time point. i Xi represents the monitoring value of the monitoring parameter at the ith time point; if the monitoring parameter is a periodic parameter, the parameter sequence X is composed of the monitoring values at the same time point in different periods in the order of the periods, and xi represents the monitoring value of the ith period at a certain time point.
[0007] Further, the abnormal monitoring values are divided into offset monitoring values, variation monitoring values and missing monitoring values, and the offset monitoring values and the variation monitoring values are extracted from the parameter sequence by using the Isolation Forest algorithm.
[0008] Further, the steps of the Isolation Forest algorithm include constructing an isolation tree, calculating an abnormal score, calculating an average abnormal score and extracting an abnormal monitoring value.
[0009] Further, the step of constructing an isolation tree is as follows: selecting a first split value a1, dividing the monitoring values in the parameter sequence into two subsets A1 and A2, which satisfy: and selecting a split value for the divided subsets and repeating the above steps until all monitoring values are divided into separate subsets A i The number of divisions experienced by a monitoring value when it is finally divided into a separate subset A i is called the path length l(x i ) of the monitoring value in the isolation tree.
[0010] Further, the formula for calculating the abnormal score s(x i ) is
[0011] where c(n) is the average path length for normalization, which is approximately calculated by the following formula:
[0012] where H(n-1) is the harmonic series, which has ; After constructing a plurality of isolation trees, the formula for calculating the average abnormal score s avg (x i ) is .
[0013] where T represents the number of constructed isolation trees, and st represents the abnormal score of the monitoring value x i calculated in the tth time.
[0014] Further, the method for extracting an abnormal monitoring value is as follows: given an abnormal judgment threshold y1 and y 2, and a time judgment threshold t0, when the monitoring value x ithe average abnormal score of the abnormal monitoring value is higher than the abnormal determination threshold y1; then the abnormal monitoring value is classified according to the following rules, (1) any isolated abnormal monitoring value is classified as a variation monitoring value; (2) for continuous abnormal monitoring values, the time difference is calculated, when the time difference is greater than the time determination threshold t0, the abnormal monitoring value is classified as a normal monitoring value, otherwise it is classified as a drift monitoring value; (3) for continuous abnormal monitoring values, the average abnormal score and the average abnormal score average value are calculated, the difference between the average abnormal score and the average abnormal score average value is calculated, and the abnormal monitoring value corresponding to the average abnormal score whose difference is greater than the abnormal determination threshold y2 is classified as a mutation monitoring value.
[0015] Further, in step S3, the correction formula for the drift monitoring value is:
[0016] is the corrected drift monitoring value, p1 is the average value of the continuous drift monitoring value, p2 is the average value of the normal monitoring value, is the drift correction value.
[0017] Further, in step S3, the correction method for the mutation monitoring value is to classify it into a missing monitoring value; that is, the average value of the two normal monitoring values closest to the missing monitoring value in time is taken as the repair value of the missing monitoring value, and the time points of the two normal monitoring values are located on the front and back sides of the time point of the missing monitoring value.
[0018] Further, in step S4, the running state of the running parameter can be divided into normal state, emergency state and fault state, and the change rate x d of the running parameter is defined as:
[0019] wherein x i is the normal monitoring value or the corrected abnormal monitoring value, x r is the rated monitoring value of the running parameter, the first deviation rate x t1 and the second deviation rate x t2 are given, when x r belongs to [x t1 , x t2 ], the running parameter is in the emergency state, and when x r belongs to (x t2 , ∞), the running parameter is in the fault state.
[0020] The application has the beneficial effect that by monitoring a certain operating parameter of the fuse, a parameter sequence of the operating parameter is obtained, then an abnormal monitoring value is extracted from the parameter sequence, and the abnormal monitoring value is corrected after classification, and finally the corrected abnormal monitoring value is used to judge the operating state of the fuse with respect to the operating parameter, the accuracy of the overall judgment is less affected by the abnormal monitoring value, and the reliability is higher. BRIEF DESCRIPTION OF DRAWINGS
[0021] Figure 1 The step flow of the operating state monitoring method of the fuse in the application. DETAILED DESCRIPTION
[0022] The technical solutions of the application will be described clearly and completely below in combination with embodiments.
[0023] As shown in the figure, it is a preferred embodiment of the application, an operating state monitoring method of a fuse. Figure 1
[0024] Step S1, determine the operating parameter of the fuse, the value obtained by monitoring the operating parameter is called the monitoring value, and the monitoring values are arranged in time sequence as the parameter sequence.
[0025] In the operation process of the fuse, the operating parameters used to judge its operating state can be roughly divided into two categories, one is the electrical parameter, such as operating voltage and operating current, the other is the non-electrical parameter, such as operating temperature and operating humidity, among which the operating voltage and operating current can be monitored by setting voltage monitoring module and current monitoring module outside the fuse respectively, and the operating temperature and operating humidity can be monitored by setting temperature sensor and humidity sensor outside the fuse respectively.
[0026] The value obtained by monitoring the above operating parameter at a certain time point is called the monitoring value of the operating parameter, and the monitoring values are arranged in turn according to the order of time points to form the parameter sequence of the operating parameter. This paper sets the parameter sequence as X, which is represented as follows:
[0027] Where x i represents the monitoring value obtained by monitoring the operating parameter at the i th time point.
[0028] In addition, when the fuse is periodically operated, the monitoring values of part of the operating parameters also present periodic changes, at this time, the operating parameters are divided into periodic parameters and non-periodic parameters, the parameter sequence of the non-periodic parameters is directly obtained by the above method, and the parameter sequence of the periodic parameters is obtained as follows: the initial time point of a given operating cycle is defined as the time difference value from the initial time, a plurality of time points are selected within the operating cycle, then the monitoring values of the periodic parameters at these time points are collected in each operating cycle, and for a plurality of monitoring values at the same time point, the monitoring values are arranged in sequence according to the operating cycle order, at this time, x i represents the monitoring value in the ith cycle at a certain time point, and for the periodic parameters, m parameter sequences are generated if there are m time points.
[0029] Step S2, the monitoring values in the parameter sequence are divided into normal monitoring values and abnormal monitoring values.
[0030] For the monitoring values, there are normal monitoring values and abnormal monitoring values, the normal monitoring values refer to those monitoring values that can be used for subsequent judgment of the operating parameter state, and the abnormal monitoring values refer to those monitoring values that cannot be used for subsequent judgment of the operating parameter state, therefore, the abnormal monitoring values need to be extracted and repaired subsequently.
[0031] For the parameter sequence, the abnormal monitoring values in the parameter sequence can be roughly divided into three types, which are offset monitoring values, variation monitoring values and missing monitoring values, wherein the offset monitoring values refer to the condition that a plurality of continuous monitoring values deviate from the normal monitoring values, the variation monitoring values refer to the condition that a single or a plurality of discontinuous monitoring values deviate from the normal monitoring values, and the missing monitoring values refer to the condition that the monitoring values are not obtained, for the condition of the missing monitoring values, since the missing monitoring values do not exist, only the time point at which the missing monitoring values are generated needs to be recorded.
[0032] For the conditions of the offset monitoring values and the variation monitoring values, the present application uses the isolated forest algorithm to extract the offset monitoring values and the variation monitoring values from the parameter sequence, and the algorithm has the characteristics that it is suitable for extracting abnormal values in a large amount of data, and the steps are as follows: Step 1, constructing an isolated tree The parameter sequence is: .
[0033] The steps of constructing the isolated tree are as follows: (1) selecting any one feature dimension of the monitoring value x i , in the present application, the monitoring value is one-dimensional, and only one dimension of the value itself exists; (2) according to the value range of the monitoring value, a first split value a1 is selected, which can be selected as the value range between the minimum value x min and the maximum value x max in the parameter sequence, or a value range is artificially given ,in yes (3) Divide the remaining monitoring values in the parameter sequence into two subsets A1 and A2, respectively satisfying: and (4) Re-split values are given within the split subsets A1 and A2, and (2), (3), and (4) are repeated until all monitoring values are split into separate subsets A1 and A2. i Within the tree, the number of splits required to ultimately divide a monitored value into a single subset is called its path length l(x) in the isolated tree. i ).
[0034] Step 2: Calculate the outlier score For each sample x i After obtaining its path length l(x) i After that, its abnormality score s(x) is calculated using the following formula. i ): .
[0035] Where c(n) is the average path length, used for normalization, and is approximately calculated using the following formula: Where H(n-1) is the harmonic series, which has .
[0036] Step 3: Calculate the average outlier score The anomaly score calculated from a single isolated tree is too random to be directly used to determine whether a monitoring value is abnormal. Steps 1 and 2 above need to be repeated: first, the first segmentation value a1 needs to be reselected, then the isolated tree needs to be constructed, and the anomaly score calculated. The average anomaly score for any monitoring value can be obtained using the following formula: Among them, s avg (x i () represents the average outlier score, and T represents the number of isolated trees constructed. This represents the result of the t-th calculation regarding the monitored value x. i Abnormal scores.
[0037] Step 4: Extract abnormal monitoring values Based on experience and the sensitivity requirements for extracting anomaly monitoring values, given anomaly determination thresholds y1 and y2... 2, And the time-based threshold t0, when the monitored value x iOnce the average abnormal score is higher than the abnormal judgment threshold y1, the monitored value can be identified as significantly different from the normal monitored value, which is an abnormal monitored value. This abnormal monitored value is not the final abnormal monitored value and needs to be classified according to the following rules: (1) Any isolated abnormal monitored value is classified as a variant monitored value. Here, isolated means that in the sequence parameters after removing missing monitored values, the monitored values at adjacent time points are all normal monitored values; (2) For continuous abnormal monitored values, here, continuous means that in the sequence parameters after removing missing monitored values, at least one of the monitored values at adjacent time points is not a normal monitored value. These continuous abnormal monitored values constitute an abnormal monitored value group, and the time difference between the time points corresponding to the first and last abnormal monitored values in the group is calculated and compared with the time judgment threshold t0. When it is greater than the time judgment threshold t0, it means that the monitored value was collected over a longer period of time. Abnormal monitoring values, at this time it can be determined that the abnormality of the monitoring values is not caused by environmental interference, but by the abnormality of the fuse itself. These continuous abnormal monitoring values are reclassified into normal monitoring values. If the time difference is less than or equal to the time judgment threshold t0, these continuous abnormal monitoring values are all classified into offset monitoring values. (3) Considering that there may also be variable monitoring values in the offset monitoring values, for the abnormal monitoring value group in step (2), regardless of whether the time difference in the group is greater than the time judgment threshold t0, it is also necessary to calculate the average abnormal score corresponding to each abnormal monitoring value in the group and the average abnormal score of all the average abnormal scores in the group, and compare the average abnormal score in the group with y2. All abnormal monitoring values corresponding to the average abnormal score of the group with an average abnormal score greater than y2 are classified into variable monitoring values. The remaining abnormal monitoring values are classified into normal monitoring values or offset monitoring values according to the method in step (2).
[0038] Step S3: Correct the abnormal monitoring values and obtain the corrected parameter sequence; Based on the status of the abnormal monitoring values, they are repaired using the following three methods.
[0039] (1) Offset monitoring value The offset monitoring value is generated because, within a certain period of time, the fuse is subjected to continuous and directional external or internal interference, causing the original normal monitoring value to shift in a specific direction. Compared with other abnormal monitoring values, the offset monitoring value can still reflect the changes in operating parameters to a certain extent within a certain period of time. It can be restored to the normal monitoring value using the following method: The average value of consecutive offset monitoring values is calculated as p1. Here, offset monitoring values with only interval variation monitoring values are considered consecutive. Multiple normal monitoring values closest to the offset monitoring value are selected, generally not exceeding the number of consecutive offset monitoring values, and the average value of these normal monitoring values is calculated as p2. For any offset monitoring value x...i , the following correction formula is used: . is the corrected offset monitoring value, which can be regarded as a normal monitoring value subsequently, wherein is the offset correction value.
[0040] (2) Variation monitoring value The variation monitoring value is caused by the internal or external instantaneous interference of the fuse. The cause of the instantaneous interference is difficult to determine, which leads to the inability to determine the variation monitoring value from the normal monitoring value by what reason. Therefore, the variation monitoring value is a meaningless monitoring value, which can be directly classified as a missing monitoring value and repaired by the same method for repairing the missing monitoring value.
[0041] (3) Missing monitoring value For the missing monitoring value, an interpolation method can be used to repair it, that is, the average of the two normal monitoring values closest to the missing monitoring value in time is taken as the repair value of the missing monitoring value, and the time points of the two normal monitoring values are located on the front and back sides of the time point of the missing monitoring value.
[0042] Step S4, classifying the running state of the running parameter, using the corrected parameter sequence to judge the current running state of the running parameter.
[0043] The running state of the running parameter can be divided into normal state, emergency state and fault state, which is usually set in a gradient manner. The specific steps are as follows, define the change rate x d of the running parameter as: Wherein x i is the normal monitoring value or the corrected abnormal monitoring value, x r is the rated monitoring value of the running parameter, the first deviation rate x t1 and the second deviation rate x t2 are given, when x r belongs to [x t1 , x t2 ], the running parameter is in an emergency state, when x r belongs to (x t2 , ∞), the running parameter is in a fault state.
[0044] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. A method of monitoring the operating state of a fuse, characterized in that: The method comprises the following steps, Step S1, determining the operating parameter of the fuse, the value obtained by monitoring the operating parameter is called the monitoring value, and the monitoring values are arranged in time sequence as a parameter sequence; Step S2, dividing the monitoring values in the parameter sequence into normal monitoring values and abnormal monitoring values; Step S3, correcting the abnormal monitoring values and obtaining a corrected parameter sequence; Step S4, classifying the operating state of the operating parameter, and using the corrected parameter sequence to judge the current operating state of the operating parameter.
2. The monitoring method of claim 1, wherein: In step S1, the operating parameter is a periodic parameter or a non-periodic parameter, and the parameter sequence of the non-periodic parameter is recorded as X, which is represented as follows, wherein x in the parameter sequence X represents a monitoring value obtained by monitoring the operating parameter at the i-th time point; if the operating parameter is a non-periodic parameter, then x i represents a monitoring value obtained by monitoring the operating parameter at the i-th time point; if the operating parameter is a periodic parameter, then the parameter sequence X is composed of monitoring values at the same time point in different periods in the order of the periods, in which case x i represents a monitoring value in the i-th period at a certain time point.
3. The method of claim 2, wherein: In step S2, the abnormal monitoring values are divided into offset monitoring values, mutation monitoring values and missing monitoring values, and the offset monitoring values and the mutation monitoring values are extracted from the parameter sequence by using the isolated forest algorithm.
4. The method of claim 3, wherein: The steps of the isolated forest algorithm include constructing an isolated tree, calculating an abnormal score, calculating an average abnormal score, and extracting an abnormal monitoring value.
5. The method of claim 4, wherein: The construction of the isolation tree proceeds as follows: a first split value a1is chosen, and the sequence of parameters is split into two subsets A1and A2, satisfying and respectively. The split value is chosen within the split subsets, and the procedure is repeated until all monitoring values are split into separate subsets A i The number of splits experienced by a monitoring value in its final split into a separate subset is called its path length in the isolation tree, l(x i ).
6. The monitoring method according to claim 5, characterized in that: Computing the abnormality score s(x i ) is given by the formula, where c(n) is the average path length, used for normalization, which is approximated by the following formula, where H(n - 1) is the harmonic series, which has ; After constructing multiple isolated trees, the average anomaly score s is calculated avg (x i ) is given by the formula, where T represents the number of constructed isolated trees, represents the anomaly score for the monitoring value x i at the tth computation.
7. The method of claim 6, wherein: The method for extracting abnormal monitoring values is as follows: given an abnormality determination threshold y1 and a time determination threshold t0, when the average abnormal score of monitoring values x 2, i is higher than the abnormality determination threshold y1, the monitoring value is an abnormal monitoring value; then, the abnormal monitoring values are classified according to the following rules in sequence: (1) any isolated abnormal monitoring value is classified as a variation monitoring value; (2) for continuous abnormal monitoring values, a time difference value is calculated, when the time difference value is greater than the time determination threshold t0, the abnormal monitoring value is classified as a normal monitoring value, otherwise, the abnormal monitoring value is classified as a drift monitoring value; (3) for continuous abnormal monitoring values, the average abnormal scores and the average abnormal score average value are calculated, the difference between each average abnormal score and the average abnormal score average value is calculated, and the abnormal monitoring value corresponding to the average abnormal score with a difference greater than the abnormality determination threshold y2 is classified as a mutation monitoring value. 8. The method of claim 7, wherein: In step S3, the correction formula for the offset monitoring value is: p1 is the average of the successive offset monitoring values, p2 is the average of the normal monitoring values, and p is the offset correction value. p1 is the average of the successive offset monitoring values, p2 is the average of the normal monitoring values, and p is the offset correction value.
9. The method of claim 8, wherein: In step S3, the correction method for the mutation monitoring value is to classify it into a missing monitoring value; that is, the two normal monitoring values closest to the missing monitoring value in time are selected, and the average value of the two normal monitoring values is taken as the repair value of the missing monitoring value, and the time points of the two normal monitoring values are located on the front and back sides of the time point of the missing monitoring value.
10. The method of claim 9, wherein: In step S4, the running state of the running parameter can be divided into a normal state, an emergency state, and a fault state, and the change rate x of the running parameter is defined as: d x = (Pn - Pn-1) / (Tn - Tn-1) where x i is the normal monitoring value or the corrected abnormal monitoring value, x r is the rated monitoring value of the operating parameter, a first deviation rate x t1 and a second deviation rate x t2 are given, when x r belongs to [x t1 , x t2 ], the operating parameter is in an emergency state, and when it belongs to (x t2 , ∞), the operating parameter is in a fault state.