A power monitoring cloud platform for a distribution system based on instrument transformers

Abstract

The present invention relates to the field of electric power monitoring technology, and specifically to a power distribution system electric power monitoring cloud platform based on a transformer, comprising: a feature acquisition module, for real-time acquisition of operating data of a distribution system, and extraction of a feature vector of the distribution system during patrol monitoring; the feature vector comprises a specific value of current and voltage, an average power of current and voltage, a phase difference of current and voltage, and an autocorrelation coefficient and a partial correlation coefficient of current and voltage; a feature detection module, for detecting operating data corresponding to the feature vector, and obtaining a prediction probability and a prediction classification corresponding to the feature vector during detection; a classification judgment module, for judging and identifying the prediction classification, and determining a detection processing method of the prediction classification when a fault occurs and a probability characteristic of the detection processing method; the reliability and stability of the distribution system are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power monitoring, and specifically to a power monitoring cloud platform for a distribution system based on current transformers. Background Art

[0002] As an important infrastructure in modern society, the power system plays a crucial supporting role in industry, commerce, and residential life. As one of the core components in the power system, the current transformer undertakes key tasks such as current measurement, protection, and control, and is of great significance to the safe and stable operation of the power system. However, there are some limitations and challenges in the traditional maintenance methods of current transformers, which lead to a series of problems in their practical applications.

[0003] For example, Chinese Patent Publication No. CN117216673A discloses a monitoring, evaluation, and maintenance platform for current transformers, which includes a multi-modal sensor module, a monitoring module, a pre-training module, a state evaluation module, a maintenance decision correction module, and a historical state database; the monitoring module is used to remotely receive the data collected by the multi-modal sensor module, and store and output the data; the historical state database stores the historical operation parameters of the current transformer and the corresponding state evaluation results; the pre-training module trains a support vector machine model based on the data in the historical state database to obtain a classifier; the state evaluation module conducts a state evaluation of the current transformer based on the collected data; the prior art adjusts the current evaluation method of the current transformer by judging the state category, but the prior art ignores the correlation between current and voltage, as well as the occurrence of corresponding faults, resulting in the lack of a judgment on the overall state in the state evaluation result after identifying the corresponding category.

[0004] For example, Chinese Patent Publication No. CN111475929A discloses an inversion verification method based on the monitoring data of a distribution network in-situ test platform. The method includes: constructing a recording system according to the actual physical topology of the in-situ test platform; obtaining the voltage data and current data of all nodes; calculating the positive sequence voltage component, positive sequence current component, zero sequence voltage component, and zero sequence current component of all nodes; calculating the positive sequence impedance and zero sequence impedance between all nodes; calculating the equivalent impedance of the π model between all nodes; constructing a digital simulation line of the in-situ test platform based on the equivalent impedance of the π model between all nodes; based on the new line impedance, conducting fault simulation and waveform reproduction through the digital simulation line of the in-situ test platform; the prior art can perform inverse verification through the components of current and voltage and impedance, and can find the corresponding magnitude value of the current, but when the current is abnormal, its period cannot be adjusted, resulting in a reduction in the overall efficiency and verification accuracy of the inverse verification. Summary of the Invention

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: A power monitoring cloud platform for a distribution system based on transformers, comprising: A feature acquisition module, configured to collect the operation data of the distribution system in real time and extract the feature vectors of the distribution system during inspection and monitoring; The feature vectors include the specific values of current and voltage, the average power of current and voltage, the phase difference of current and voltage, the autocorrelation coefficient and the partial correlation coefficient of current and voltage.

[0006] A feature detection module, configured to detect the operation data corresponding to the feature vectors and obtain the prediction probability and prediction classification corresponding to the feature vectors during detection.

[0007] A classification and judgment module, configured to judge and identify the prediction classification, and determine the detection processing method and the probability characteristics of the detection processing method when a fault occurs in the prediction classification.

[0008] A period adjustment module, configured to verify the detection processing method, determine the monitoring period and the interval period selected for each detection processing method, and adjust the size of the monitoring period according to the corresponding probability characteristics of the detection processing method.

[0009] A correction and maintenance module, configured to correct the abnormal fluctuations of each detection processing method during a fault, determine the correction category of the fault after correction, and repair the equipment according to the correction category.

[0010] The beneficial effects of the present invention are as follows: Through the real-time acquisition and feature extraction functions, the system of the present invention can discover and process potential faults in a timely manner, improving the monitoring efficiency; By comprehensively using a variety of feature vectors for prediction classification, the accuracy and reliability of fault prediction are improved; Dynamically adjusting the monitoring period according to the system state and fault probability characteristics reduces the monitoring cost and improves the monitoring efficiency; By correcting the abnormal fluctuations of each detection processing method during a fault and determining the correction category, a scientific basis is provided for equipment maintenance, reducing the maintenance cost and downtime; The reliability and stability of the distribution system are improved, ensuring the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The present invention will be further described below with reference to the drawings and embodiments.

[0012] Figure 1 It is a system framework diagram of a power monitoring cloud platform for a distribution system based on transformers.

[0013] Figure 2 It is a system flow chart of a power monitoring cloud platform for a distribution system based on transformers.

[0014] Figure 3 It is a schematic flow chart of prediction classification in a feature detection module of a power monitoring cloud platform for a distribution system based on transformers.

[0015] Figure 4 It is a schematic flowchart for obtaining a detection and processing method in a classification and judgment module of a power monitoring cloud platform for a distribution system based on a mutual inductor. Specific embodiments

[0016] The embodiments of the present invention will be described in detail below. The following described embodiments are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention. For those not specified in the embodiments regarding specific technologies or conditions, they shall be carried out according to the technologies or conditions described in the literature in the field or according to the product instructions.

[0017] Refer to Figure 1 、 Figure 2 A power monitoring cloud platform for a distribution system based on a mutual inductor, comprising: a feature acquisition module, a feature detection module, a classification and judgment module, a cycle adjustment module, and a correction and maintenance module.

[0018] The feature acquisition module collects the operation data of the distribution system in real time, extracts the feature vectors, and transmits the extracted feature vectors to the feature detection module; the feature detection module detects the operation data corresponding to the feature vectors, obtains the prediction probability and prediction classification, and transmits the prediction probability and prediction classification to the classification and judgment module; the classification and judgment module judges and identifies the prediction classification, determines the detection and processing method and its probability characteristics, and transmits the determined detection and processing method and its probability characteristics to the cycle adjustment module; the cycle adjustment module verifies the detection and processing method, determines the monitoring cycle and the interval cycle, and adjusts the monitoring cycle according to the probability characteristics, and feeds back the adjusted monitoring cycle and interval cycle to the feature acquisition module for data acquisition according to the new cycle; the correction and maintenance module corrects the abnormal fluctuations of each detection and processing method during a fault, determines the correction category, and performs maintenance according to the correction category, providing maintenance suggestions and / or directly controlling the maintenance equipment to operate.

[0019] The feature acquisition module is used to collect the operation data of the distribution system in real time and extract the feature vectors of the distribution system during inspection and monitoring.

[0020] The feature detection module is used to detect the operation data corresponding to the feature vectors and obtain the prediction probability and prediction classification corresponding to the feature vectors during detection.

[0021] The classification and judgment module is used to judge and identify the prediction classification and determine the detection and processing method and the probability characteristics of the detection and processing method when a fault occurs.

[0022] The cycle adjustment module is used to verify the detection and processing method, determine the monitoring cycle and the interval cycle selected for each detection and processing method, and adjust the size of the monitoring cycle according to the corresponding probability characteristics of the detection and processing method.

[0023] The correction and maintenance module is used to correct the abnormal fluctuations of various detection and processing methods during faults, determine the correction category of the fault after correction, and repair the equipment according to the correction category.

[0024] The operating data obtained here includes the current and voltage data during the operation of the power distribution system, and the current and voltage data are used to extract feature vectors according to the current waveform and voltage waveform to obtain the corresponding changes in current and voltage during the inspection and monitoring of the power distribution system.

[0025] The main purpose is to regularly inspect and maintain the power distribution system to prevent corresponding equipment failures, identify possible situations after identifying the corresponding feature vectors, and express this situation in a predictive manner, such as using predictive classification and prediction probability to obtain possible faults or other problems under the current feature vectors; and determine the processing methods that can be used at this time, and which detection and processing methods are more suitable for processing, and at the same time adjust the monitoring period each time to adjust the monitoring frequency at this time; this monitoring period also needs to be used to process those parts that trip or are regarded as abnormal but ineffective after inspection, and these will indicate possible problems, but monitor unclear faults, so as to complete the correction of faults or problems; finally, it is to correct these specific faults that occur and complete the maintenance and monitoring of the platform according to the corresponding correction category.

[0026] For example, obtain the center frequency point and signal bandwidth of the operating data, and calculate the average power and phase difference of the operating data according to the center frequency point and signal bandwidth, and use the average power and phase difference as the characteristic information of the feature vector.

[0027] At the same time, it is necessary to calculate the autocorrelation coefficient and partial correlation coefficient of the operating data within the corresponding time, and use the autocorrelation coefficient and partial correlation coefficient as the characteristic information of the feature vector.

[0028] Therefore, the output feature vector at this time includes the specific values of current and voltage, the average power of current and voltage, the phase difference of current and voltage, the autocorrelation coefficient and partial correlation coefficient of current and voltage.

[0029] The center frequency point represents the frequency of the power grid, such as 50Hz or 60Hz; the signal bandwidth represents the range of frequency components in the current and voltage, and this value can be obtained by a spectrum analyzer to obtain the center frequency point and signal bandwidth of the corresponding current and voltage.

[0030] For the autocorrelation coefficient and partial correlation coefficient, the collected current and voltage data are arranged in the form of a time series, and the setting method of the lag stage is used to determine the autocorrelation coefficient and partial correlation coefficient of the current and voltage at this time.

[0031] For example, the autocorrelation coefficient can be expressed as follows. First, the running data is converted into a time series, a lag phase is set for the time series, and the autocorrelation coefficient is calculated based on the lag phase.

[0032] ; where represents the autocorrelation coefficient when the lag phase is k, represents the value of the running data at time t, and this value represents the value corresponding to the selected time t in the time series of current and voltage, represents the value of the running data at time t - k, represents the average value of the running data, and at this time, it represents the average value of the corresponding current and voltage in the time series, represents the length of the time series, and the value range of t is from 1 to n, represents the lag phase.

[0033] For the partial correlation coefficient, after calculating the autocorrelation coefficient, a recursive calculation is performed to obtain the partial correlation coefficient.

[0034] ; where represents the partial correlation coefficient when the lag phase is k, represents the autocorrelation coefficient when the lag phase is k, represents the autocorrelation coefficient when the lag phase is k - j, represents the partial correlation coefficient when the lag phase is j, represents the lag phase, and the value range of j is from 1 to k - 1; , where represents the partial correlation coefficient when the lag phase is 1, the autocorrelation coefficient when the lag phase is 1.

[0035] After calculating the autocorrelation coefficient here, the part greater than the initial threshold is selected as the feature vector for subsequent processing. The initial threshold is set to 0.05. When the autocorrelation coefficient and partial correlation coefficient are greater than 0.05, it indicates a significant change. When it is greater than 0.2, it indicates that the current and voltage at this time conform to the normal change; the purpose of selecting 0.05 is to identify that there is an abnormality in the current distribution system but not to show the specific situation.

[0036] In an embodiment of the present invention, a feature detection module is used to detect the running data corresponding to the feature vector and obtain the prediction probability and prediction classification corresponding to the feature vector during detection.

[0037] The prediction probability and prediction category of the feature vector are the output results in the classification task, and they respectively represent the confidence level of a certain feature vector belonging to each category and the final classification result.

[0038] The predicted probability obtained at this time represents the probability value of the currently obtained feature vector with respect to each classification. When obtaining the predicted probability at this time, spectral feature extraction will first be performed on the feature vector, and autocorrelation analysis and partial correlation analysis will be performed on the spectral features, so as to obtain the finally output predicted classification.

[0039] Such as Figure 3 As shown, the way to obtain the predicted classification is as follows: classify the feature information in the feature vector in sequence, and sequentially obtain the first classification information corresponding to the specific values of current and voltage, average power, and phase difference; the first classification information is inclined to obtain the operating conditions of current and voltage in the overall process, such as whether the current and voltage are stable, whether there are obvious fault problems, and whether the phase difference of the current and voltage is normal, affected by the inductance in the current transformer at this time, etc., to quantify the state of the power distribution system at this time.

[0040] Obtain the second classification information corresponding to the autocorrelation coefficient and partial correlation coefficient of current and voltage; the second classification information is used to better capture the structure of current and voltage with respect to time variation, to determine which situation the current and voltage tend to change, so as to obtain a more accurate predicted probability.

[0041] Under normal circumstances, to obtain the predicted probability and predicted classification of the feature vector, by comparing the feature information in the feature vector with the data set in the historical data, the confidence level of the feature information is determined, and the part with the maximum value of the confidence level is selected as the predicted classification according to the confidence level, and this confidence level is also regarded as the predicted probability.

[0042] In the present invention, when obtaining the confidence level of the feature information, the feature information will be combined to obtain a comprehensive confidence level, and the comprehensive confidence level will be classified to obtain the classification most corresponding to the current situation, and the probability values of the first classification information and the second classification information will be calculated respectively.

[0043] For example, the way to obtain the probability value of the first classification information is as follows: calculate the fitting probability value corresponding to the specific values of current and voltage, calculate the power probability value corresponding to the average power of current and voltage, calculate the phase probability value corresponding to the phase difference of current and voltage, and take the comprehensive value of the fitting probability value, power probability value, and phase probability value as the probability value of the first classification information.

[0044] The fitting probability value is obtained according to the change of current and voltage in the corresponding time series, by obtaining the slope value corresponding to current and voltage, and calculating the distribution of the slope value, to obtain the fitting probability value corresponding to current and voltage.

[0045] The slope value corresponding to current and voltage is expressed as follows: ; where represents the slope value corresponding to current and voltage, represents the value of the operating data at time t, represents the length of the time series, and the value range of t is from 1 to n; the calculated slope value is used to identify whether the current-voltage shows a stable trend or a scattered trend under normal operating conditions, so as to identify whether the power distribution system is operating normally at this time.

[0046] The fitting probability value is expressed as follows. Obtain the slope value corresponding to the current-voltage, and calculate the fitting slope value.

[0047] ; where represents the fitting probability value, represents the slope value corresponding to the current-voltage, represents the exponential constant.

[0048] The power probability value is expressed as follows. Obtain the average power corresponding to the current-voltage, and calculate the power probability value.

[0049] ; where represents the power probability value, represents the average power corresponding to the current-voltage, 、 represents the cross-validation parameter, and the set values are 0.1 and -1 in turn.

[0050] The phase probability value is expressed as the ratio of the phase difference between the current-voltage and the corresponding standard phase difference in the historical data; the standard phase difference is the average value of the phase differences in the historical data.

[0051] The probability value of the first classification information is expressed as: ; where represents the probability value of the first classification information, represents the phase probability value; represents the probability constant. At this time, the probability constant is expressed as a value between 0 and 1. The value of the probability constant represents the joint probability value of the slope value, average power, and phase difference corresponding to the current current-voltage appearing in the historical data. The joint probability value is different from the corresponding values of the fitting probability value, power probability value, and phase probability value calculated above, so as to quantify whether the final classification obtained during the comprehensive classification at this time can meet the corresponding requirements. These values obtained by comprehensively considering multiple conditions during classification, for example, the joint probability value tends to consider the number of occurrences of the corresponding value in the overall population, while the fitting probability value and power probability value tend to consider the probability distribution of the corresponding value, and the phase probability value tends to consider the ratio of the corresponding value to the standard value. These values will represent the comprehensive situation under different conditions and will be classified in turn according to the identified conditions to obtain more specific sub-classification categories under each classification.

[0052] The second classification information is obtained based on the probability values of the autocorrelation coefficient and partial autocorrelation coefficient of the current and voltage. The probability values of the autocorrelation coefficient and partial autocorrelation coefficient are used as the probability values of the second classification information. The probability values of the autocorrelation coefficient and partial autocorrelation coefficient are expressed as the ratio of the number of occurrences of the corresponding autocorrelation coefficient and partial autocorrelation coefficient values in the historical data to the total number of times; the probability value of the second classification information is the product of the probability values of the autocorrelation coefficient and partial autocorrelation coefficient.

[0053] Calculate the probability values of the first classification information and the second classification information respectively, combine the probability values of the first classification information and the second classification information, and select the classification with the largest combined probability value as the predicted classification for output.

[0054] The probability value obtained after combining the first classification information and the second classification information here is the predicted probability corresponding to the feature vector.

[0055] The way to combine the probability values of the first classification information and the second classification information is to perform weighted averaging on the first classification information and the second classification information, and the weights used are 0.7 and 0.3 in sequence.

[0056] When obtaining the combined probability value, multiple probability values will be obtained according to the different classifications selected at this time, and the largest probability value is selected as the predicted classification for output at this time.

[0057] In an embodiment of the present invention, a classification judgment module is used to judge and identify the predicted classification, and determine the detection and processing method and the probability characteristics of the detection and processing method when a failure occurs in the predicted classification.

[0058] At this time, the detection and processing method refers to the processing measures adopted for the predicted classification after obtaining a certain predicted classification, and the corresponding program flow; for example, when identifying the predicted classification corresponding to problems such as tripping, abnormal current fluctuation, and abnormal current consumption, the detection and processing method at this time is to alarm, record the log, and adjust the corresponding current and voltage.

[0059] The acquisition method of the detection and processing method is to determine the fault attributes corresponding to each predicted classification according to the obtained predicted classification. The fault attributes include fault type, fault cause, and processing suggestions.

[0060] Map the predicted classification to the fault attributes to generate a category attribute comparison table. The category attribute comparison table is used to map the information contained in the predicted classification into a comparison table to represent the possible fault problems under the current predicted classification.

[0061] Analyze the fault attributes in sequence according to the category attribute comparison table to determine the detection and processing method under the predicted classification.

[0062] For example, the category attribute comparison table can be as follows.

[0063] Prediction classification, normal; Fault type, none; Fault cause, none; Handling suggestion, no handling required.

[0064] Prediction classification, overload; Fault type, overload; Fault cause, excessive current; Handling suggestion, check the load and reduce the current.

[0065] Prediction classification, short circuit; Fault type, short circuit; Fault cause, circuit short circuit; Handling suggestion, disconnect the power supply and check the circuit.

[0066] Prediction classification, grounding; Fault type, grounding; Fault cause, poor grounding; Handling suggestion, check the grounding wire to ensure good grounding.

[0067] When obtaining the category attribute comparison table, it is possible to clearly know the fault attributes corresponding to the current prediction classification. At the same time, since the prediction classification represents the overall relative state, it is necessary to verify whether the data within the prediction classification completely matches the fault attributes to know the detection and handling methods to be adopted.

[0068] As Figure 4 shown, the implementation method of determining the detection and handling method under the prediction classification by sequentially analyzing the fault attributes according to the category attribute comparison table can be as follows.

[0069] Construct a first analysis pair equal in number to the number of fault attributes from the category attribute comparison table. The first analysis pair is used to generate multiple analysis pairs from the attribute comparison table according to the number of fault attributes. Each first analysis pair contains a fault attribute and the set rating of the fault attribute. The set rating represents the matching degree between the current fault attribute and the prediction classification.

[0070] The matching degree between the fault attribute and the prediction classification is determined by dividing using the occurrence time of the fault, determining the duration and scope of influence of the fault in the fault attribute, and setting the matching degree between the fault attribute and the prediction classification according to the duration and scope of influence of the fault.

[0071] For example, the duration of the fault is expressed as the ratio of the duration length value to the average duration length of the corresponding fault attribute. This ratio is denoted as the index value of the duration of the fault; the scope of influence of the fault is expressed as the ratio of the number of system components affected by the fault to the total number of all system components, and this ratio is regarded as the index value of the scope of influence of the fault; the two are weighted and summed to obtain the matching degree between the fault attribute and the prediction classification.

[0072] ; where represents the matching degree between the fault attribute and the prediction classification, represents the index value of the duration of the current fault, The index value representing the scope of influence of the current fault The minimum value of the index value representing the duration of the fault The minimum value of the index value representing the scope of influence of the fault The maximum value of the index value representing the duration of the fault The maximum value of the index value representing the scope of influence of the fault The weight representing the duration of the fault The weight representing the scope of influence of the fault Represents an exponential factor, which is a constant greater than 1 and is used to amplify the influence of the current duration and scope of influence. For example, the exponential factor can be taken as 1.4 or the maximum value among the ratios of the index values of the current fault duration and scope of influence to the minimum index value can be selected as the exponential factor at this time; the maximum and minimum values selected at this time are obtained by looking up the corresponding fault attributes in the historical data, and the weights used can be set to 0.6 and 0.4 in the order of duration and scope of influence.

[0073] Generate a second analysis pair according to the fault type from the category attribute comparison table. The number of the second analysis pairs is the same as the number of existing fault types. The second analysis pair includes the fault type and the risk probability, and the second analysis pair is used to analyze the risk probability existing under different fault types, that is, the probability of occurrence of this fault type under the content included in the category attribute comparison table, so as to verify the verification degree of the fault occurring at this time.

[0074] The risk probability is the ratio of the number of times the current fault type occurs to the total number of faults, or the ratio of the number of times the current fault type occurs to the total number of faults occurring in the current category attribute comparison table. Both of these values can reflect the relative situation of the current fault when it appears.

[0075] Generate a third analysis pair according to the fault cause from the category attribute comparison table. The third analysis pair includes the fault cause and the fault classification probability. The fault classification probability represents the probability that the fault cause is classified into the corresponding fault type. The fault classification probability will be stored in the database. When the category attribute comparison table is mapped to the corresponding fault cause, this fault classification probability will be obtained at the same time to determine the probability of the classification in which the fault cause is located at this time, so as to quantify the specific situation of the fault at this time.

[0076] Compare the first analysis pair, the second analysis pair, and the third analysis pair with the data of the fault attributes, fault types, and fault causes in the historical data in turn, and calculate the fault support degree and fault confidence degree of the first analysis pair, the second analysis pair, and the third analysis pair respectively. When both the fault support degree and the fault confidence degree are greater than the preset threshold, use the detection and processing methods existing in the historical data as the output detection and processing methods.

[0077] The fault support degree and fault confidence degree here represent the support degree and confidence degree of normal calculation, and the preset thresholds are both set to 0.6 to obtain a list of detection and processing methods that meet the preset thresholds; that is, it is required that for each of the three analysis pairs when calculating the fault support degree and fault confidence degree, the support degree and confidence degree of each analysis pair reach the preset threshold before corresponding historical data will be selected, thereby obtaining the corresponding detection and processing methods.

[0078] After outputting the detection and processing methods, it is also necessary to determine the probability characteristics of the detection and processing methods at this time. The probability characteristics are expressed as the duration, influence range, and occurrence frequency of the fault. The duration and influence range of the fault are obtained through the first analysis pair in the category attribute comparison table, and the occurrence frequency is obtained based on the number of occurrences of the fault within a unit time. For example, the unit time is set to 1 week to obtain the occurrence frequency of the corresponding fault.

[0079] In an embodiment of the present invention, the cycle adjustment module is used to verify the detection and processing methods, determine the monitoring cycle and interval cycle selected for each detection and processing method, and adjust the size of the monitoring cycle according to the corresponding probability characteristics of the detection and processing methods.

[0080] The detection cycle and interval cycle at this time are directly obtained through the detection and processing methods. These cycles are preset and will be included in the detection and processing methods. For example, when the values of the fault frequency, duration, and influence range are detected to be too large or too small, the monitoring cycle is adjusted to dynamically adjust the monitoring cycle.

[0081] The monitoring cycle refers to the time interval from the start of monitoring to the next monitoring. In other words, it is the time length of each monitoring operation. The monitoring cycle determines how often the system will perform a complete data collection and analysis.

[0082] The interval cycle refers to the frequency or interval time of data collection within the monitoring cycle. It determines how often the system collects data within the monitoring cycle.

[0083] Here, it mainly focuses on adjusting the monitoring cycle, and different interval cycles are adjusted. The interval cycle only affects the collection interval, while the monitoring cycle affects the overall processing and analysis. By adjusting the monitoring cycle, the overall processing interval and prediction preparation can be synchronously adjusted, thereby improving the sensitivity of power monitoring.

[0084] For example, the method of adjusting the size of the monitoring cycle is as follows.

[0085] ; where represents the adjusted monitoring cycle, represents the initial value of the monitoring cycle, Represents an exponential constant, An index value representing the fault frequency of the current fault. This index value represents the normalized fault frequency, An index value representing the duration of the current fault, An index value representing the scope of influence of the current fault, The average value of the index values representing the fault frequency of the current fault, The average value of the index values representing the duration of the current fault, The average value of the index values representing the scope of influence of the current fault; the average values of the corresponding index values are all obtained by extracting the average value from historical data.

[0086] The monitored period of the final output is represented as a numerical value of a time length, indicating how often a complete data analysis and processing is performed.

[0087] In an embodiment of the present invention, a correction and maintenance module is used to correct the abnormal fluctuations of each detection and processing method during a fault, determine the correction category of the corrected fault, and perform maintenance on the equipment according to the correction category.

[0088] Here, it is used to correct the current and voltage during a fault. The correction method is based on the obtained detection and processing method. At the same time, the correction category represents the fault type after the fusion processing of the detection and processing method, indicating the repair result required by the current operating state of the power distribution system. For example, the correction category represents correcting the identified fault type and fault cause, adding new fault descriptions, or adjusting and optimizing the divided fault types, so that the system can accurately identify and discover new fault types.

[0089] The method for correcting the abnormal fluctuations of each detection and processing method during a fault is to splice the detection and processing methods, select the one with the largest index value in the detection and processing methods as the splicing center, superimpose each detection and processing method with the splicing center, and output the superimposed splicing center as the maximum splicing center.

[0090] The index values in the detection and processing methods here include the setting rating of the fault attribute, the fault type and risk probability, the fault cause and fault classification probability, and the values of the corresponding steps when the detection and processing method is implemented. At this time, the splicing method is to select the one with the largest comprehensive index value as the splicing center, and perform the superimposition processing in the way of weighted averaging all the detection and processing methods in turn. The superimposed splicing center will be a more accurate, more robust or more reliable data representation, which can be used for subsequent decision-making, analysis or further processing.

[0091] After obtaining the maximum splicing center, the corrected category is extracted based on the maximum splicing center. At this time, the index value of the maximum splicing center is selected, and the maximum value of the fault type in the maximum splicing center is used as the corrected category at this time.

[0092] Finally, the overall is corrected according to the obtained corrected category to obtain the final corrected category, and the equipment is repaired according to the obtained corrected category.

[0093] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention, and still be covered by the protection scope of the present invention.

Claims

1. A power distribution system power monitoring cloud platform based on transformer, characterized in that: include: The feature acquisition module is used to collect the operation data of the distribution system in real time and extract the feature vector of the distribution system during inspection and monitoring; the feature vector includes the specific value of current and voltage, the average power of current and voltage, the phase difference of current and voltage, the autocorrelation coefficient and partial correlation coefficient of current and voltage; A feature detection module is used to detect the operating data corresponding to the feature vector and obtain the prediction probability and prediction classification corresponding to the feature vector during the detection; The classification judgment module is used to judge and identify the predicted classification, determine the detection and processing method of the predicted classification when a fault occurs, and the probability characteristics of the detection and processing method; According to the category attribute comparison table, the fault attributes are analyzed in turn, and the detection and processing methods under the prediction classification are determined as follows: The category attribute comparison table is used to construct first analysis pairs equal to the number of fault attributes, each first analysis pair includes a fault attribute and a setting rating of the fault attribute, and the setting rating represents the matching degree between the current fault attribute and the predicted classification; The category attribute comparison table generates second analysis pairs according to the fault type, the number of the second analysis pairs is consistent with the number of existing fault types, and the second analysis pairs include the fault type and the risk probability; The category attribute comparison table generates a third analysis pair according to the fault cause, the third analysis pair including the fault cause and the fault classification probability; The first analysis pair, the second analysis pair, and the third analysis pair are compared with the data of fault attributes, fault types, and fault causes in the historical data in turn, and the fault support and fault confidence of the first analysis pair, the second analysis pair, and the third analysis pair are calculated respectively. When the fault support and the fault confidence are both greater than the preset threshold, the detection processing method existing in the historical data is used as the output detection processing method; The cycle adjustment module is used to verify the detection processing method, determine the monitoring cycle and interval cycle selected by each detection processing method, and adjust the monitoring cycle size according to the corresponding probability characteristics of the detection processing method; The correction and maintenance module is used to correct the abnormal fluctuations of each detection and processing method when a fault occurs, determine the correction category of the corrected fault, and repair the equipment according to the correction category.

2. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 1 is characterized in that: The method of obtaining the prediction classification is as follows: classifying the characteristic information in the characteristic vector in sequence, and obtaining the specific value of the current and voltage, the average power and the first classification information corresponding to the phase difference in sequence; Obtaining second classification information corresponding to the autocorrelation coefficient and partial correlation coefficient of current and voltage; The probability values ​​of the first classification information and the second classification information are calculated respectively, the probability values ​​of the first classification information and the second classification information are combined, and the classification with the largest probability value after the combination is selected as the output prediction classification.

3. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 2 is characterized in that: The probability value of the first classification information is obtained as follows: Calculate the fitting probability value corresponding to the specific value of the current and voltage, calculate the power probability value corresponding to the average power of the current and voltage, and calculate the phase probability value corresponding to the phase difference of the current and voltage, and use the comprehensive value of the fitting probability value, the power probability value and the phase probability value as the probability value of the first classification information; The probability value of the second classification information is the product of the probability values ​​of the autocorrelation coefficient and the partial correlation coefficient.

4. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 3 is characterized in that: The fitting probability value is expressed as follows. The slope value corresponding to the current and voltage is obtained, and the fitting slope value is calculated: ； in, represents the fitting probability value, Indicates the slope value corresponding to the current and voltage, represents the exponential constant; The power probability value is expressed as follows: Get the average power corresponding to the current and voltage, and calculate the power probability value: ； in, represents the power probability value, represents the average power corresponding to the current and voltage, , represents the cross validation parameters; The phase probability value is expressed as the ratio of the phase difference between the current and voltage to the corresponding standard phase difference in the historical data; The probability value of the first classification information is expressed as: ； in, represents the probability value of the first classification information, represents the phase probability value; represents a probability constant.

5. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 1 is characterized in that: The detection and processing method is obtained by determining the fault attributes corresponding to each prediction classification according to the obtained prediction classification, and the fault attributes include fault type, fault cause, and processing suggestion; Map the predicted classification with the fault attributes to generate a category attribute comparison table; According to the category attribute comparison table, the fault attributes are analyzed in turn to determine the detection and processing method under the predicted classification.

6. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 1 is characterized in that: The matching degree between fault attributes and predicted classification is expressed as: ； in, represents the matching degree between the fault attribute and the predicted classification, An indicator value indicating the duration of the current fault. Indicates the index value of the impact range of the current fault. The minimum value of the indicator indicating the duration of the fault. The minimum value of the indicator value indicating the impact range of the fault, The maximum value of the indicator value representing the duration of the fault, The maximum value of the indicator value indicating the impact range of the fault. represents the weight of the fault duration, The weight representing the impact range of the fault, Represents the exponential factor.

7. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 6 is characterized in that: Probabilistic characteristics are expressed as the duration, impact scope and frequency of occurrence of the fault.

8. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 6 is characterized in that: The way to adjust the monitoring period size is as follows: ； in, represents the adjusted monitoring period, Indicates the initial value of the monitoring period, represents the exponential constant, The index value indicating the fault frequency of the current fault, The average value of the indicator value representing the fault frequency of the current fault, The average value of the indicator value representing the duration of the current fault, Indicates the average value of the indicator value of the impact range of the current fault.

9. The power distribution system power monitoring cloud platform based on mutual inductor according to claim 1, characterized in that: The method for correcting the abnormal fluctuation of each detection and processing method during a fault is: The detection processing methods are spliced, the one with the largest index value among the detection processing methods is selected as the splicing center, each detection processing method is superimposed with the splicing center, and the superimposed splicing center is output as the maximum splicing center; The maximum value of the fault type in the largest splicing center is taken as the correction category at this time.

Images