A multi-source equipment operation fault monitoring system for a power distribution service recorder

Abstract

The present application relates to the technical field of fault supervision, in particular to a multi-source equipment operation fault supervision system for power supply and distribution service recorder, the system comprises a processor and a memory, the processor executes the computer program stored in the memory to realize the following steps: obtaining the state conflict representation value between any two states at the current time, obtaining the target state conflict index value corresponding to each state combination in the to-be-screened state combination set at the current time according to the probability value and the state conflict representation value corresponding to any two states at the current time, screening the to-be-screened state combination set according to the target state conflict index value to obtain the target state combination set, and obtaining the supervision result of the to-be-supervised equipment at the current time by using D-S evidence theory according to the target state combination set and the probability value corresponding to each state of the to-be-supervised equipment at the current time. The present application can improve the effect of supervising the operation fault of the power supply and distribution system equipment.

Description

Technical Field

[0001] This invention relates to the field of fault monitoring technology, specifically to a multi-source equipment operation fault monitoring system for power supply and distribution service recorders. Background Technology

[0002] Currently, to ensure power supply continuity and reduce operation and maintenance costs, it is usually necessary to monitor the operational faults of equipment in the power supply and distribution system. Existing technologies typically begin by collecting multi-source data from operating equipment using power supply and distribution service recorders. This collected data is then input into a random forest, which outputs the equipment state and its probability value under different data categories. The output state probability value is then used as the Basic Trust Assignment (BPA) in the Data Structures (DS) evidence theory, and multi-source fusion analysis is performed to ultimately obtain a comprehensive and reliable judgment of the equipment's operating status or a monitoring result of the equipment's operation. Furthermore, existing DS evidence theories generally establish an identification framework based on a known set of equipment states before fusion analysis. This means that current DS fusion theories require the identification framework to include all known possible state combinations. However, the operating status of power supply and distribution system equipment is a non-stationary, multi-dimensional, and dynamic process that is affected by load changes, environmental changes, equipment aging, and topology adjustments. The constant changes in the state structure cause the coupling relationships between states to vary with the environment or operating conditions. This can lead to conflicts in certain state combinations at certain times or indistinguishable combinations at specific periods. Statistically, these conflicting state combinations cannot provide a stable basis for confidence assignment in the DS evidence theory. If these conflicting state combinations are retained during fusion, it will not only result in longer fusion calculation times and higher computational complexity, but also lead to fuzzy assignment of basic trust values, continuous conflicts or false trust accumulation during fusion calculations, thereby affecting the timeliness of monitoring equipment malfunctions and causing distortion of the final fusion result. In other words, the current method of determining the identification framework leads to poor monitoring of equipment malfunctions in power supply and distribution systems. Therefore, how to adaptively adjust the identification framework, that is, adaptively adjust the set of state combinations, to improve the effectiveness of monitoring equipment malfunctions in power supply and distribution systems has become an urgent problem to be solved. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a multi-source equipment operation fault monitoring system for power supply and distribution service recorders, the specific technical solution of which is as follows:

[0004] One embodiment of the present invention provides a multi-source equipment operation fault monitoring system for a power supply and distribution service recorder, including a processor and a memory, wherein the processor executes a computer program stored in the memory to perform the following steps:

[0005] Obtain the set of state combinations to be screened, the probability values ​​of each state of the device to be monitored at different times, the operating data of each state, and the final output state of each operating data of the device to be monitored at different times.

[0006] Based on the running data corresponding to each state at each moment in the preset current nearest time period and the final output state corresponding to each running data, the feature value between any two states at the current moment is obtained. The feature value includes the difference feature value of running data change trend and the historical joint occurrence probability feature value. Based on the feature value between any two states at the current moment, the state conflict characterization value between any two states at the current moment is obtained. Based on the probability value corresponding to any two states at the current moment and the state conflict characterization value, the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current moment is obtained.

[0007] The set of states to be screened is filtered according to the target state conflict index value to obtain the target state combination set. Based on the target state combination set and the probability values ​​corresponding to each state of the device to be monitored at the current time, the monitoring result of the device to be monitored at the current time is obtained using DS evidence theory.

[0008] Beneficial Effects: This invention first obtains a set of state combinations to be screened, probability values ​​of each state of the device to be monitored at different times, operational data of each state, and the final output state of each operational data of the device to be monitored at different times. Then, based on the operational data of each state at each time in the preset current nearest time period and the final output state of each operational data, it obtains the feature value between any two states at the current time. The feature value includes the feature value of the difference in the trend of operational data change and the feature value of the probability of historical joint occurrence. Based on the feature value between any two states at the current time, it obtains the state conflict characterization value between any two states at the current time. Based on the probability value and the state conflict characterization value of any two states at the current time, it obtains the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time. Finally, it screens the set of state combinations to be screened based on the target state conflict index value to obtain the target state combination set. Based on the target state combination set and the probability value of each state of the device to be monitored at the current time, it uses DS evidence theory to obtain the monitoring result of the device to be monitored at the current time. Furthermore, this invention, based on the target state conflict index value corresponding to the state combination, comes from the adaptive adjustment identification framework or state combination set. This not only reduces the fusion calculation time and computational complexity, but also avoids problems such as fuzzy assignment in basic trust allocation, continuous conflict or false trust accumulation in fusion calculation, thereby improving the effectiveness of monitoring the operation faults of power supply and distribution system equipment. Attached Figure Description

[0009] 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.

[0010] Figure 1 This is a flowchart of a multi-source equipment operation fault monitoring method for a power supply and distribution service recorder according to the present invention. Detailed Implementation

[0011] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0012] 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.

[0013] This embodiment provides a multi-source equipment operation fault monitoring system for a power supply and distribution service recorder, including a processor and a memory. The processor executes a computer program stored in the memory to implement a method for monitoring multi-source equipment operation faults in a power supply and distribution service recorder, such as... Figure 1 As shown, the multi-source equipment operation fault monitoring method of the power supply and distribution service recorder includes the following steps:

[0014] Step S001: Obtain the set of state combinations to be screened, the probability values ​​of each state of the device to be monitored at different times, the operating data of each state, and the final output state of each operating data of the device to be monitored at different times.

[0015] Currently, before performing fusion analysis, DS algorithms typically establish an identification framework based on a known set of device states. This means existing DS fusion theories require the identification framework to encompass all known possible state combinations. However, the operating status of power distribution system equipment is a non-stationary, multi-dimensional, dynamic process that constantly changes with load variations, environmental changes, equipment aging, and topology adjustments. This causes the coupling relationships between device states to change with the environment or operating conditions, resulting in some state combinations conflicting at certain times or becoming indistinguishable at specific periods. Statistically, these conflicting state combinations cannot provide a stable confidence assignment basis for DS evidence theory. Therefore, retaining these conflicting state combinations during fusion not only leads to longer fusion calculation times but also increases computational complexity. High complexity can lead to fuzzy assignment of basic trust values, continuous conflicts or false trust accumulation during fusion calculations, thus affecting the timeliness of equipment operation fault monitoring and causing distortion of the final fusion result. Distorted fusion results result in lower accuracy and reliability of monitoring. In other words, the traditional method of determining the identification framework leads to poor monitoring of the operating status or faults of power supply and distribution system equipment. To ensure or improve the effectiveness of monitoring the operating status or faults of power supply and distribution system equipment, this embodiment will adaptively adjust the identification framework or state combination set based on the target state conflict index value corresponding to each state combination at different times. The target state conflict index value reflects the degree of conflict between states in the state combination, thereby improving the effectiveness of monitoring the operating status or faults of power supply and distribution system equipment. Furthermore, since power supply and distribution systems contain many devices, and the method for monitoring the operation faults of any power supply and distribution system device is the same, this embodiment will describe the operation fault monitoring process of any power supply and distribution system device as an example for ease of understanding, and will refer to it as the device to be monitored.

[0016] After identifying the equipment to be monitored, this embodiment uses a power supply and distribution service recorder to collect different types of operational data of the equipment at different times, obtaining multi-source operational data of the equipment at different times. This means obtaining various operational data of the equipment at different times. The power supply and distribution service recorder can simultaneously collect multiple types of operational data. At any given time, the operational data collected by the power supply and distribution service recorder includes, but is not limited to, the current, voltage, temperature, and vibration of the equipment. Thus, the current, voltage, temperature, and vibration of the equipment at different times can be obtained. Alternatively, as another real-time method, the data collected by the power supply and distribution service recorder at different times and then preprocessed can be used as multi-source operational data of the equipment at different times. The data preprocessing in this embodiment includes missing value supplementation and extreme outlier replacement. In this embodiment, missing values ​​can be supplemented using linear interpolation, and extreme outliers can be marked using box plots and replaced with previous values. The preprocessing process is a known technique.

[0017] After obtaining the operational data of the monitored equipment at different times, each type of operational data is used as an independent evidence source and input into the random forest. Each evidence source outputs a final state and the probability value of each state corresponding to each evidence source after passing through the random forest. In other words, based on the random forest, the final output state corresponding to each operational data at different times and the probability value corresponding to each state can be obtained. The specific acquisition process is as follows:

[0018] First, the state category corresponding to each type of operational data is obtained. In this embodiment, the state categories are only abnormal and normal. That is, for any operational data category, its corresponding state category is only normal or abnormal. For example, if an operational data category is current, then the state categories corresponding to that operational data category are only abnormal current and normal current. Then, a random forest corresponding to different operational data categories is constructed. The state output by the random forest corresponding to any operational data category is only normal or abnormal. For example, if an operational data type is current, then the state categories output by the random forest corresponding to that operational data category are only abnormal current and normal current. The training process of the random forest is a well-known technique. Then, the operational data of the device to be monitored at different times are input into the random forest corresponding to the corresponding operational data category, and the final output state corresponding to each operational data at different times and the probability value corresponding to each state are output. For example, for the current at any time t, the current at time t is input into the random forest corresponding to the current. The final output state of the current at time t, as well as the probability values ​​of the current being abnormal and normal at time t, can be output. If the probability value of the current being normal is the largest at this time, then the final output state of the current at time t is the current being normal. The determination method for the final output state of other operating data at time t and the probability values ​​corresponding to different states is the same as the determination method for the final output state of the current at time t and the probability values ​​corresponding to different states.

[0019] After obtaining the final output state and probability values ​​corresponding to each operational data at different times, the operational data corresponding to each state of the monitored device at different times is obtained. That is, for any state at any time, if the state belongs to the state category corresponding to operational data category c, then the operational data belonging to operational data category c among all operational data of the monitored device at that time is the operational data corresponding to that state at that time. Specifically, for a certain operational data c at any time, operational data c is the operational data of the monitored device at that time. If the probability value of the output state after operational data c is input into the random forest corresponding to the operational data type to which operational data c belongs includes the probability value of state C, then the operational data corresponding to state C at the corresponding time is determined to be operational data c. For example, for the current anomaly at time t, the operational data corresponding to the current anomaly at time t refers to the current of the monitored device at time t obtained above or the current of the monitored device at time t obtained above.

[0020] Then, obtain the set of state combinations to be screened: First, denote the set of all state categories corresponding to all the operating data categories input to the random task as the state set of the device to be monitored. For example, if the operating data categories input to the random task include current, voltage, temperature, and vibration, then the state set consists of normal current, normal voltage, normal temperature, normal vibration, abnormal current, abnormal voltage, abnormal temperature, and abnormal vibration. Then, perform non-repeating combinations on the states in the state set, and denote the set of state combinations obtained as the initial state combination set. The number of states in the state combinations in the initial state combination set is greater than 2 and less than or equal to A, where A is the number of states in the state set. That is, first select two different states from the state set for non-repeating combination, then select three different states from the state set for non-repeating combination, and so on, until A different states are selected from the state set for non-repeating combination and then stop. Denote the set of all combinations obtained as the initial state combination set. Due to the physical constraints of the power supply and distribution system, some state combinations are mutually exclusive or incompatible in actual operation. These physically incompatible state combinations can be directly eliminated during the identification framework construction stage without relying on data analysis. That is, in order to reduce the amount of computation and ensure the regulatory effect, this embodiment first obtains the mutually exclusive state combinations in the initial state combination set, and then eliminates all mutually exclusive state combinations in the initial state combination set. The new set of state combinations remaining after elimination is denoted as the state combination set to be screened. And this is only a preliminary screening. In order to further ensure the regulatory effect, the state combination set to be screened needs to be screened again. The process of determining whether a state combination is a mutually exclusive state combination is as follows: for any state combination, if there are all state categories corresponding to any operating data category in the state combination, then the state combination is determined to be a mutually exclusive state combination. For example, if a state combination consists of current abnormality and current normality, and current abnormality and current normality cannot be established or occur at the same time, then the state combination is determined to be a mutually exclusive state combination.

[0021] Therefore, this embodiment can obtain the set of state combinations to be screened, the probability values ​​of each state of the device to be monitored at different times, the operating data of each state, and the final output state of each operating data of the device to be monitored at different times through the above process. The probability values ​​of each state, the operating data of each state, and the final output state of each operating data are the data support for further screening of the set of state combinations to be screened.

[0022] Step S002: Based on the running data corresponding to each state at each moment in the preset current nearest time period and the final output state corresponding to each running data, obtain the feature value between any two states at the current moment. The feature value includes the running data change trend difference feature value and the historical joint occurrence probability feature value. Based on the feature value between any two states at the current moment, obtain the state conflict characterization value between any two states at the current moment. Based on the probability value corresponding to any two states at the current moment and the state conflict characterization value, obtain the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current moment.

[0023] The above-mentioned method of obtaining the set of state combinations to be screened only removes mutually exclusive combinations, but does not remove conflicting state combinations. As the environment or operating conditions change, conflicting state combinations will still exist in the set of state combinations to be screened at different times. This embodiment then needs to filter out conflicting state combinations from the set of state combinations to be screened at different times. The basis for this filtering is the target state conflict index value. That is, this embodiment needs to obtain the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time. The target state conflict index value reflects the severity of the conflict between states in a state combination. The specific process for obtaining the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time is as follows:

[0024] Since the number of states in each state combination is greater than or equal to 2, and the combination of three or more states is essentially a supplement and extension of the pairwise combinations, but does not change the coupling relationship between the two combinations themselves, and the physical mechanism of the operational failure of the monitored equipment is usually directly related to the pairwise states, such as when the monitored equipment has a certain type of abnormality, it will cause the monitored equipment to have another type of abnormality through physical connection or information interaction, so this embodiment will use the pairwise states as the most basic unit to analyze and obtain the target state conflict index value corresponding to each state combination in the set of state combinations to be screened.

[0025] First, a preset current nearest neighbor time period is obtained. This preset current nearest neighbor time period includes the current time. In specific applications, the implementer needs to set the preset current nearest neighbor time period based on actual conditions such as the collection interval of similar adjacent operational data. Furthermore, the collection interval of similar adjacent operational data must be much shorter than the length of the preset current nearest neighbor time period. For example, in this embodiment, the time period consisting of 10 minutes before the current time and the current time can be used as the preset current nearest neighbor time period. Then, based on the operational data corresponding to each state at each time in the preset current nearest neighbor time period and the final output state corresponding to each operational data, the characteristic value of the difference in operational data change trends between any two states at the current time is obtained. The historical joint occurrence probability feature value, the operational data change trend difference feature value, and the historical joint occurrence probability feature value are all feature values. The operational data change trend difference feature value can reflect the possibility of conflict between two states, and the historical joint occurrence probability feature value can also reflect the possibility of two states co-occurring or the possibility of conflict between two states. The operational data change trend difference feature value and the historical joint occurrence probability feature value are key parameters for subsequently determining the conflict index value of the target state. Therefore, the specific process of obtaining the operational data change trend difference feature value and the historical joint occurrence probability feature value between any two states at the current moment is as follows: For state a and state b:

[0026] First, the time series consisting of the running data corresponding to state 'a' at all times in the current nearest neighbor time period is denoted as the first sequence, and the time series consisting of the running data corresponding to state 'b' at all times in the current nearest neighbor time period is denoted as the second sequence. For example, if state 'a' belongs to the state category corresponding to current, then the time series consisting of the current of the monitored device acquired at all times in the current nearest neighbor time period is the first sequence. Then, based on the differences between adjacent data in the first and second sequences and the differences in the number of maximum values ​​between the first and second sequences, the characteristic value of the difference in the trend of running data change between any two states at the current time is obtained. Finally, based on the final output state corresponding to the running data in the first and second sequences, the characteristic value of the historical joint occurrence probability between any two states at the current time is obtained.

[0027] In this embodiment, the specific process for obtaining the characteristic value of the difference in the trend of running data change between any two states at the current moment, based on the difference between adjacent data in the first sequence and the second sequence and the difference in the number of maxima between the first sequence and the second sequence, is as follows:

[0028] A first rate of change sequence is obtained based on the differences between adjacent data in the first sequence, and a second rate of change sequence is obtained based on the differences between adjacent data in the second sequence. The d-th data point in the first rate of change sequence is the ratio of the (d+1)-th data point minus the d-th data point in the first sequence to the d-th data point itself. The expression for the d-th data point in the first rate of change sequence is: The f-th data point in the second rate of change sequence is the ratio of the (f+1)-th data point minus the f-th data point in the second sequence to the f-th data point itself. The expression for the f-th data point in the second rate of change sequence is: , For the d-th data in the first sequence, For the (d+1)th data in the first sequence, For the f-th data in the second sequence, For the (f+1)th data point in the second sequence, the absolute values ​​of the differences between corresponding data points in the first and second rate of change sequences are accumulated and then normalized, and this is recorded as the first characterization value. The maxima in the first and second sequences are obtained, and the normalized result of the absolute value of the difference between the number of maxima in the first and second sequences is calculated and recorded as the second characterization value. Here, the hyperbolic tangent function is used for normalization. The mean of the first and second characterization values ​​is calculated and used as the characteristic value of the difference in the trend of running data changes between state a and state b at the current moment. The specific expression for the characteristic value of the difference in the trend of running data changes between state a and state b at the current moment is:

[0029]

[0030] in, Let be the characteristic value of the difference in the trend of running data changes between state a and state b at the current moment, tanh() be the hyperbolic tangent function, and J be the total number of data in the first rate of change sequence and also the total number of data in the second rate of change sequence. For the j-th data point in the first rate of change sequence, Let N1 be the j-th data point in the second rate of change sequence, N2 be the number of maxima in the first sequence, and N1 be the number of maxima in the second sequence. Since the operating data of the equipment directly reflects its actual operating behavior, and when the operating data corresponding to two states show a large difference in changing trends, it indicates that the two states are driven by completely different mechanisms, and these mechanisms are difficult to satisfy simultaneously within the same time period. This results in the two states exhibiting characteristics of being mutually exclusive or rarely coexisting in real-world operating conditions. The greater the likelihood of conflict between two states with mutually exclusive or rarely coexisting characteristics, the more likely conflict exists. The larger and A larger value indicates greater inconsistency in the magnitude or direction of changes in the operational data corresponding to state a and state b within the same time period; more significant structural differences in the operational data corresponding to state a and state b; and greater inconsistency in the volatility of changes in the operational data corresponding to state a and state b. This suggests a greater difference in the changing trends of the operational data corresponding to state a and state b within the preset current nearest time period. The larger and When it is larger, The larger, therefore when A larger value indicates a greater difference in the changing trends of the operational data corresponding to state a and state b in the current nearest time period. It also indicates that states a and b are more likely to be mutually exclusive or rarely coexist in real-world conditions, or that there is a greater possibility of conflict between states a and b at the current moment. Conversely, a smaller value indicates a greater difference in the changing trends of the operational data corresponding to state a and state b in the current nearest time period. The smaller the value, the smaller the difference in the trend of the running data corresponding to state a and state b in the current nearest time period. It also indicates that the possibility of a conflict between state a and state b at the current moment is smaller.

[0031] In this embodiment, the specific process of obtaining the historical joint occurrence probability feature value between any two states at the current moment based on the final output state corresponding to the running data in the first sequence and the second sequence is as follows:

[0032] Based on the comparison between the final output state corresponding to each running data in the first sequence and state a, each running data in the first sequence is marked, and the marked value of each running data in the first sequence is obtained. Based on the comparison between the final output state corresponding to each running data in the second sequence and state b, each running data in the second sequence is marked, and the marked value of each running data in the second sequence is obtained. The specific process for obtaining the marked value of the running data is as follows: Determine whether the final output state corresponding to the g-th running data in the first sequence is state a. If yes, mark the g-th running data in the first sequence as 1, that is, the marked value of the g-th running data in the first sequence is 1 at this time; otherwise, mark the g-th running data in the first sequence as 0, that is, the marked value of the g-th running data in the first sequence is 0 at this time. Determine whether the final output state corresponding to the h-th running data in the second sequence is state b. If yes, mark the h-th running data in the second sequence as 1, that is, the marked value of the h-th running data in the second sequence is 0 at this time. The h-th running data in the sequence is marked as 1; otherwise, the h-th running data in the second sequence is marked as 0. The sequence formed by the marked values ​​of the running data in the first sequence is denoted as the first marked value sequence, and the sequence formed by the marked values ​​of the running data in the second sequence is denoted as the second marked value sequence. The i-th marked value in the first marked value sequence is the marked value of the i-th running data in the first sequence. Similarly, the second marked value sequence is calculated by multiplying the corresponding positions in the first and second marked value sequences, and the result is denoted as the comprehensive sequence. That is, the k-th data in the comprehensive sequence is the product of the k-th data in the first and second marked value sequences. The mean of all data in the comprehensive sequence is calculated and used as the historical joint occurrence probability feature value between state a and state b at the current time. The specific calculation expression for the historical joint occurrence probability feature value between state a and state b at the current time is:

[0033]

[0034] in, Let M be the historical joint occurrence probability feature value between state a and state b at the current moment, and M be the total number of data in the comprehensive sequence. Let m be the m-th data point in the composite sequence. The more non-zero data points in the composite sequence, the greater the probability that states a and b frequently occur simultaneously in the near future, and the greater the probability that states a and b coexist in the near future. However, the more non-zero data points in the composite sequence... The larger, It is also larger, so when A larger value indicates a higher probability that states a and b coexist at the current moment, and a lower probability that there is a conflict between states a and b at the current moment. Conversely, a smaller value indicates a lower probability that there is a conflict between states a and b at the current moment. The smaller the value, the lower the probability that state a and state b coexist at the current moment, and the greater the possibility that there is a conflict between state a and state b at the current moment.

[0035] Since the characteristic values ​​of the difference in the trend of operational data changes and the historical joint occurrence probability characteristic values ​​can reflect the likelihood of conflict between states, this embodiment, after obtaining the characteristic values ​​of the difference in the trend of operational data changes and the historical joint occurrence probability characteristic values, obtains the state conflict characterization value between any two states at the current moment based on the characteristic values ​​of the difference in the trend of operational data changes and the historical joint occurrence probability characteristic values ​​between any two states at the current moment. Subsequently, the target state conflict index value will be determined based on the state conflict characterization value. The specific process of obtaining the state conflict characterization value between any two states at the current moment is as follows:

[0036] For states a and b: Calculate the mean of the negative correlation mapping result of the historical joint occurrence probability feature value between states a and b at the current time and the difference feature value of the operational data change trend between states a and b at the current time. Use this as the state conflict characterization value between states a and b at the current time. Here, the negative correlation mapping is achieved by subtracting the historical joint occurrence probability feature value from a constant of 1. The expression for calculating the state conflict characterization value between states a and b at the current time is as follows: The larger the state conflict representation value between state a and state b at the current moment, the smaller the probability of state a and state b coexisting at the current moment, and the greater the possibility of a conflict between state a and state b at the current moment.

[0037] Since a set of states may contain multiple pairwise combinations, not every pairwise combination has an equal impact on the set. To further characterize the degree of conflict within the entire set, this embodiment assigns weights to the state conflict representation values ​​of each pairwise state. This ensures that the most critical pairwise states have a greater influence on the overall conflict level of the set, meaning the most critical pairwise states dominate the target state conflict index value. Furthermore, since the BPA value of a state typically represents the basic confidence level of the evidence supporting that state, the BPA reflects the importance of each state within the set. A higher BPA indicates greater importance or coreity within the set. Therefore, this embodiment will next use the BPA values ​​of the pairwise states of a state combination to... The proportion of BPA values ​​for all pairwise states in the set is used as the final weight value for the state conflict representation value of the corresponding pairwise states. Subsequently, the obtained state conflict representation values ​​are weighted based on this weight value to obtain the target state conflict index value. The BPA value of the aforementioned state is the probability value corresponding to that state. Therefore, in this embodiment, the state conflict representation value between any two states at the current time will be weighted according to the probability values ​​corresponding to any two states at the current time to obtain the target state conflict index value for each state combination in the set of state combinations to be screened at the current time. The specific process of weighting the state conflict representation value between any two states at the current time according to the probability values ​​corresponding to any two states at the current time to obtain the target state conflict index value for each state combination in the set of state combinations to be screened at the current time is as follows: For any state combination R:

[0038] For each state in the state combination R, perform pairwise combinations without repetition, and denote all resulting combinations as sub-combinations corresponding to state combination R. Obtain the initial BPA value of each sub-combination corresponding to state combination R at the current time. The sum of the probability values ​​corresponding to the two states in any sub-combination at the current time is the initial BPA value of that sub-combination at the current time. That is, if the two states in a sub-combination are state a and state b, then the sum of the probability values ​​corresponding to state a and state b at the current time is the initial BPA value of that sub-combination at the current time. Calculate the cumulative sum of the initial weight values ​​of all sub-combinations corresponding to state combination R at the current time, and denote it as the comprehensive BPA value of state combination R at the current time. Denote the ratio of the initial BPA value of each sub-combination corresponding to state combination R at the current time to the comprehensive BPA value as the weight value of the corresponding sub-combination at the current time. The expression for the weight value of the v-th sub-combination corresponding to state combination R at the current time is: ,in, Let V be the initial BPA value of the v-th sub-combination corresponding to state combination R at the current time, and V be the number of sub-combinations corresponding to state combination R at the current time. The larger the value, the more important or core the v-th sub-combination is among all sub-combinations corresponding to state combination R. The weighted state conflict representation value of the corresponding sub-combination is the product of the state conflict representation value between two states in each sub-combination corresponding to state combination R at the current time and the weight value of the corresponding sub-combination. If the two states in the v-th sub-combination corresponding to state combination R at the current time are state a and state b, then the expression for the weighted state conflict representation value of the v-th sub-combination corresponding to state combination R at the current time is: ,in, Let v be the weight value of the v-th sub-combination. Let $\frac{ ...

[0039] Therefore, this embodiment can obtain the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time through the above process.

[0040] Step S003: Filter the set of state combinations to be screened according to the target state conflict index value to obtain the target state combination set. Based on the target state combination set and the probability values ​​corresponding to each state of the device to be monitored at the current time, use DS evidence theory to obtain the monitoring result of the device to be monitored at the current time.

[0041] Since the target state conflict index value reflects the severity of conflict between states in a state combination, this embodiment, after obtaining the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time, filters the set of state combinations to be screened based on the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time, obtaining the target state combination set at the current time. The target state conflict index value is the identification framework of the DS evidence theory at the current time, and the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time is used to filter the set of state combinations to be screened. The specific process for filtering to obtain the target state combination set at the current moment is as follows: In the set of state combinations to be filtered, state combinations whose target state conflict index value is greater than a preset conflict threshold are removed. The new set of all remaining state combinations after removal is denoted as the target state combination set at the current moment. That is, if the target state conflict index value corresponding to a certain state combination at the current moment is greater than the preset conflict threshold, it indicates that the state combination is a conflicting state combination, and this state combination will cause the final fusion result to not converge. Therefore, this state combination is removed from the set of state combinations to be filtered, and only the remaining state combination set is allowed to participate in subsequent fusion calculations. In specific applications, implementers need to set the preset conflict threshold according to the value range of the target state conflict index value, experimental statistics, and other actual conditions. For example, in this embodiment, 0.5 can be selected as the preset conflict threshold.

[0042] After obtaining the target state combination set at the current moment, the monitoring result of the monitored device at the current moment is obtained by using the DS evidence theory based on the target state combination set at the current moment and the probability values ​​of each state of the monitored device at the current moment. Based on the output monitoring result, it can be determined whether the monitored device has an operational fault at the current moment. If the output is an abnormal state, then it is determined that the monitored device has an operational fault at the current moment. That is, by combining the target state combination set at the current moment and the probability values ​​corresponding to each state of the monitored device at the current moment, the monitoring result of the monitored device at the current moment can be generated or output through the DS evidence theory. The known recognition framework, that is, the target state combination set, and the process of fusing multi-source evidence, that is, the probability values ​​corresponding to each state of the monitored device at the current moment, to finally output the monitoring result, is a well-known technology.

[0043] Thus, this embodiment completes the monitoring of operational faults in power supply and distribution system equipment. Furthermore, by filtering out high-conflict state combinations, this embodiment effectively solves the problem of distorted fusion results caused by high-conflict state combinations in the identification framework, ensuring smooth convergence of the final fusion result. When making fault judgments based on the fusion results, it can improve the reliability and stability of operational fault monitoring. In other words, this embodiment adaptively adjusts the identification framework or state combination set based on the target state conflict index value corresponding to the state combination. This not only reduces fusion calculation time and computational complexity but also avoids problems such as fuzzy assignment in basic trust allocation, continuous conflict or false trust accumulation during fusion calculation. As a result, it can improve the effectiveness of monitoring the operational status or operational faults of power supply and distribution system equipment.

[0044] In summary, this embodiment first obtains the set of state combinations to be screened, the probability values ​​of each state of the device to be monitored at different times, the operating data of each state, and the final output state of each operating data of the device to be monitored at different times. Then, based on the operating data of each state at each time in the preset current nearest time period and the final output state of each operating data, the feature value between any two states at the current time is obtained. The feature value includes the difference feature value of the changing trend of operating data and the historical joint occurrence probability feature value. Based on the feature value between any two states at the current time, the state conflict characterization value between any two states at the current time is obtained. Based on the probability value and the state conflict characterization value of any two states at the current time, the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current time is obtained. Finally, the set of state combinations to be screened is screened according to the target state conflict index value to obtain the target state combination set. Based on the target state combination set and the probability values ​​of each state of the device to be monitored at the current time, the monitoring result of the device to be monitored at the current time is obtained using DS evidence theory. Furthermore, this embodiment is based on the target state conflict index value corresponding to the state combination, which comes from the adaptive adjustment identification framework or the state combination set. This not only reduces the fusion calculation time and computational complexity, but also avoids problems such as fuzzy assignment in basic trust allocation, continuous conflict or false trust accumulation in fusion calculation, thereby improving the effectiveness of monitoring the operating status or operating faults of power supply and distribution system equipment.

[0045] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 this application, and should all be included within the protection scope of this application.

Claims

1. A multi-source equipment operation fault monitoring system for a power supply and distribution service recorder, comprising a processor and a memory, characterized in that, The processor executes the computer program stored in the memory to perform the following steps: Obtain the set of state combinations to be screened, the probability values ​​of each state of the device to be monitored at different times, the operating data of each state, and the final output state of each operating data of the device to be monitored at different times. Based on the running data corresponding to each state at each moment in the preset current nearest time period and the final output state corresponding to each running data, the feature value between any two states at the current moment is obtained. The feature value includes the difference feature value of running data change trend and the historical joint occurrence probability feature value. Based on the feature value between any two states at the current moment, the state conflict characterization value between any two states at the current moment is obtained. Based on the probability value corresponding to any two states at the current moment and the state conflict characterization value, the target state conflict index value corresponding to each state combination in the set of state combinations to be screened at the current moment is obtained. The set of states to be screened is screened according to the target state conflict index value to obtain the target state combination set. Based on the target state combination set and the probability values ​​corresponding to each state of the device to be monitored at the current time, the monitoring result of the device to be monitored at the current time is obtained using the DS evidence theory. The method for obtaining the state conflict representation value between any two states at the current moment includes: taking the mean of the negative correlation mapping result of the historical joint occurrence probability feature value between state a and state b at the current moment and the difference feature value of the change trend of the running data between state a and state b at the current moment as the state conflict representation value between state a and state b at the current moment.

2. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 1, characterized in that, The method for obtaining the set of state combinations to be filtered includes: The set of all types of states of the device to be monitored is denoted as the state set. The states in the state set are combined without repetition, and the set of the combined states is denoted as the initial state combination set. The number of states in the state combination of the initial state combination set is greater than 2 and less than or equal to A, where A is the number of states in the state set. The mutually exclusive state combinations in the initial state combination set are removed, and the new set of the remaining state combinations after removal is denoted as the state combination set to be screened. The combination of states that cannot exist at the same time is a mutually exclusive state combination.

3. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 2, characterized in that, The probability values ​​corresponding to each state of the monitored equipment at different times and the final output state corresponding to each operating data are obtained by inputting the multi-source operating data of the monitored equipment collected by the power supply and distribution service recorder at the corresponding time into the random forest.

4. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 1, characterized in that, The method for obtaining the running data corresponding to each state includes: If the probability value of the state output by the random forest after the operational data c of the monitored device collected at any time is input into the random forest includes the probability value of state C, then the operational data corresponding to state C at that time is determined to be operational data c.

5. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 1, characterized in that, Methods for obtaining the feature values ​​between any two states at the current time include: For states a and b: the time series consisting of the running data corresponding to state a at all times in the current nearest neighbor time period is denoted as the first sequence, and the time series consisting of the running data corresponding to state b at all times in the current nearest neighbor time period is denoted as the second sequence; based on the differences between adjacent data in the first and second sequences and the differences in the number of maxima between the first and second sequences, the characteristic value of the difference in the changing trend of running data between any two states at the current time is obtained; based on the final output state corresponding to the running data in the first and second sequences, the characteristic value of the historical joint occurrence probability between any two states at the current time is obtained.

6. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 5, characterized in that, Methods for obtaining the feature values ​​of the difference in the trend of changes in running data between any two states at the current moment include: Based on the differences between adjacent data in the first and second sequences, a first rate of change sequence and a second rate of change sequence are obtained respectively. The d-th data in the first rate of change sequence is the ratio of the result of subtracting the d-th data in the first sequence from the (d+1)-th data in the first sequence to the d-th data. The f-th data in the second rate of change sequence is the ratio of the result of subtracting the f-th data in the second sequence from the (f+1)-th data in the second sequence to the f-th data. The absolute values ​​of the differences between the corresponding position data in the first and second rate of change sequences are accumulated and then normalized, and recorded as the first characterization value. The normalized result of the absolute value of the difference between the number of maxima in the first sequence and the number of maxima in the second sequence is denoted as the second characterization value; The average of the first and second characterization values ​​is used as the characteristic value of the difference in the trend of change of running data between state a and state b at the current moment.

7. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 5, characterized in that, The method for obtaining the historical joint occurrence probability feature value between any two states at the current time includes: Determine whether the final output state corresponding to the g-th running data in the first sequence is state a. If yes, then record the flag value of the g-th running data in the first sequence as 1; otherwise, record the flag value of the g-th running data in the first sequence as 0. Determine whether the final output state corresponding to the h-th running data in the second sequence is state b. If yes, record the flag value of the h-th running data in the second sequence as 1; otherwise, record the flag value of the h-th running data in the second sequence as 0. The sequence formed by the label values ​​of the running data in the first sequence is denoted as the first label value sequence, the sequence formed by the label values ​​of the running data in the second sequence is denoted as the second label value sequence, the product of the first label value sequence and the second label value sequence is denoted as the comprehensive sequence, and the mean of all data in the comprehensive sequence is used as the historical joint occurrence probability feature value between state a and state b at the current time.

8. The multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 1, characterized in that, The method for obtaining the target state conflict index value includes: For any state combination: perform non-repeating pairwise combinations on all states in the state combination, and record all combinations as sub-combinations corresponding to the state combination. Record the sum of the probability values ​​of the two states in the sub-combination at the current time as the initial BPA value of the corresponding sub-combination at the current time. Record the sum of the initial BPA values ​​of all sub-combinations corresponding to the state combination at the current time as the comprehensive BPA value of the state combination at the current time. Record the ratio of the initial BPA value to the comprehensive BPA value of the sub-combination at the current time as the weight value of the corresponding sub-combination at the current time. Record the product of the state conflict characterization value between the two states in the sub-combination at the current time and the weight value of the corresponding sub-combination as the weighted state conflict characterization value of the corresponding sub-combination at the current time. Record the sum of the weighted state conflict characterization values ​​of all sub-combinations corresponding to the state combination as the target state conflict index value of the state combination at the current time.

9. A multi-source equipment operation fault monitoring system for a power supply and distribution service recorder as described in claim 1, characterized in that, Methods for obtaining the target state combination set include: In the set of state combinations to be screened, state combinations whose target state conflict index value is greater than a preset conflict threshold are removed, and the new set of all remaining state combinations after removal is denoted as the target state combination set.

Images