Power transformation equipment working condition detection system and method based on self-adaption

Abstract

The invention discloses a power transformation equipment working condition detection system and method based on self-adaption, and relates to the field of power equipment state monitoring. The method comprises the following steps: collecting multi-source monitoring data of power transformation equipment and carrying out standardization processing to generate standard time sequence data; dividing a time window based on working conditions, extracting characteristic parameters, and constructing a characteristic track; establishing and updating an adaptive reference library containing standard reference points and health tolerance boundaries under different working conditions based on historical health data; comparing the real-time characteristic track with a standard reference point and a health tolerance boundary of a corresponding working condition in a self-adaptive reference library, calculating a deviation degree and generating an operation health index; and determining a dynamic early warning threshold value according to the historical operation health index data, and judging the state of the equipment by comparing the real-time operation health index with the dynamic early warning threshold value and a preset rigid alarm threshold value. And early discovery of weak degradation of the internal coordination relation of the power transformation equipment is realized by constructing a comparison mechanism of a self-adaptive health benchmark and a feature trajectory.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring, specifically to an adaptive substation operating condition monitoring system and method. Background Technology

[0002] With the deepening of smart grid construction, condition monitoring and health management of substation equipment have become crucial for ensuring the safe and stable operation of the power system. Currently, the condition monitoring of substation equipment mainly relies on the real-time acquisition and monitoring of multi-source parameters such as temperature, current, and vibration. Existing detection methods generally employ single-parameter or multi-parameter limit-exceeding alarm mechanisms based on fixed thresholds. These methods set safe operating thresholds for various parameters, triggering an alarm when the monitored data exceeds the threshold. In addition, some more advanced methods assess equipment status by analyzing the spectral characteristics or trend changes of specific parameters. These technical solutions can, to a certain extent, identify obvious equipment anomalies or faults.

[0003] However, under complex operating conditions, power equipment involves the coupling of multiple physical processes, including electrical, thermal, and mechanical processes. Independent analysis of a single parameter is insufficient to comprehensively reflect the overall health status of the equipment. Especially for early, subtle, latent faults, their manifestation is often not a significant change in the absolute value of a single parameter, but rather a gradual deterioration of the coordination relationship between multiple state parameters. Existing methods have limited ability to identify such "latent pattern anomalies." Furthermore, the operating status of equipment changes dynamically with factors such as load and environment, and individual equipment also exhibits differences. Fixed thresholds or models are insufficient for accurate adaptive judgment. Therefore, there is an urgent need to develop an intelligent detection method that can comprehensively assess the coordination relationship of multiple parameters and adapt to changes in individual equipment and operating conditions, in order to achieve more sensitive perception and early warning of early abnormal states of equipment. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide an adaptive substation equipment operating condition detection system and method to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an adaptive method for detecting the operating conditions of power equipment, comprising:

[0006] S1: Collect multi-source monitoring data of power equipment and perform standardization processing to generate standard time series data;

[0007] S2: Based on the time windows of standard time series data divided by working conditions, feature parameters are extracted in each window and feature trajectories are constructed in chronological order;

[0008] S3: Based on historical health data, establish and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions;

[0009] S4: Compare the real-time feature trajectory with the standard reference points and health tolerance boundaries of the corresponding working conditions in the adaptive reference library, calculate the deviation, and generate the operating health index.

[0010] S5: Determine the dynamic early warning threshold based on historical data of the operating health index, and determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

[0011] The present invention is further configured such that S1 includes:

[0012] Using the data stream with the highest sampling frequency among the multi-source monitoring data as the time reference, the remaining data streams are uniformly aligned to the timestamp sequence of the reference through an interpolation algorithm;

[0013] A statistical discrimination method based on a sliding time window is used to identify and replace transient abnormal data points in each monitoring data sequence;

[0014] Identify and mark the time periods with consecutive missing data in each monitoring data sequence;

[0015] The multi-source monitoring data, after time alignment, noise filtering, and missing data labeling, are combined according to a preset dimensional order to generate a standard time-series data vector.

[0016] The present invention is further configured such that the multi-source monitoring data includes electrical quantity monitoring data, thermal quantity monitoring data, mechanical quantity monitoring data, and operating condition label data, wherein the electrical quantity monitoring data includes the effective value of the current and active power on the high-voltage side of the transformer, the thermal quantity monitoring data includes the top oil temperature and the winding hot spot temperature, the mechanical quantity monitoring data includes the effective value of the core vibration acceleration, and the operating condition label data includes load rate range labels used to characterize the operating conditions of the equipment.

[0017] The present invention is further configured such that S2 includes:

[0018] Based on the operating condition label data, periods of continuous load rate ranges with a duration greater than the preset stable threshold are divided into stable operating condition windows.

[0019] Within each stable operating window, based on standard time-series data, the electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters are calculated.

[0020] The electrothermal conversion characteristic parameters are obtained by calculating the slope of the linear regression between the active power and the rate of change of the winding hot spot temperature within the window.

[0021] The heat transfer characteristic parameters are obtained by calculating the time delay corresponding to the maximum value of the cross-correlation function between the winding hot spot temperature and the top oil temperature within the window.

[0022] The electromechanical coupling characteristic parameters are obtained by fitting a quadratic function relationship between the effective value of vibration acceleration and the effective value of current within the window, and by calculating the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient.

[0023] The electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters calculated for each steady-state operating window are combined to form a characteristic point;

[0024] The feature points corresponding to each steady-state operating condition window are arranged in chronological order to construct the feature trajectory.

[0025] The present invention is further configured such that S3 includes:

[0026] Based on the load rate interval corresponding to the generation of each feature point in the feature trajectory, the feature points corresponding to the historical health data are classified into different load rate intervals to form multiple operating condition subsets.

[0027] For each subset of working conditions, calculate the mean vector and standard deviation vector of all feature points in each feature parameter dimension, determine the mean vector as the standard reference point under that working condition, and determine the health tolerance boundary under that working condition based on the mean vector and standard deviation vector.

[0028] Establish a mapping relationship between load rate range, standard reference point and health tolerance boundary, and store it in the form of structured data to complete the construction of adaptive reference library;

[0029] During real-time monitoring, new feature points that are identified as characterizing the health status of equipment through independent status assessment are assigned to the corresponding operating condition subset according to their load rate range. The standard reference point and health tolerance boundary under the updated operating condition subset are then recalculated to complete the update of the adaptive reference library.

[0030] The present invention is further configured such that S4 includes:

[0031] Based on the load rate range corresponding to each feature point in the real-time feature trajectory, the corresponding standard reference point and health tolerance boundary are matched and obtained from the adaptive reference library;

[0032] Based on the numerical deviation of each feature point from its standard reference point in each feature parameter dimension, and the threshold interval width of the health tolerance boundary in each feature parameter dimension, the normalized deviation of each feature point is calculated.

[0033] Calculate the arithmetic mean of the normalized deviations of all feature points in the real-time feature trajectory, and use it as the average deviation of the trajectory.

[0034] Calculate the standard deviation of the normalized deviation of all feature points relative to the average deviation, and use it as the deviation fluctuation of the trajectory.

[0035] The average deviation is added to the deviation fluctuation weighted by a preset weighting coefficient, and the sum is mapped by a monotonically decreasing function to generate the operational health index.

[0036] The present invention is further configured such that calculating the normalized deviation of each feature point specifically includes:

[0037] Calculate the difference between the value of the feature point in each dimension of the feature parameter and the value of its standard reference point in the corresponding dimension;

[0038] Divide the difference by the threshold interval width of the health tolerance boundary of the corresponding feature parameter dimension to obtain the normalized component of that dimension.

[0039] The normalized deviation of a feature point is obtained by summing the squares of the normalized components of all feature parameter dimensions and then taking the square root of the sum.

[0040] The present invention is further configured such that S5 includes:

[0041] Based on a sliding window of a preset time length, moving statistics are performed on the historical operational health index series to calculate the moving mean and moving standard deviation of the historical operational health index within the window.

[0042] The dynamic warning threshold is obtained by subtracting the moving standard deviation from the moving mean and multiplying it by a preset sensitivity coefficient.

[0043] The latest operational health index is compared with both the dynamic early warning threshold and the preset rigid alarm threshold, and the equipment status is determined based on the comparison results.

[0044] When the operating health index is greater than or equal to the dynamic early warning threshold, the equipment is determined to be in normal condition;

[0045] When the operating health index is less than the dynamic early warning threshold and the operating health index is greater than or equal to the preset rigid alarm threshold, the device is determined to be in an early warning state.

[0046] When the operating health index is lower than the preset rigid alarm threshold, the device is determined to be in an alarm state.

[0047] The present invention is further configured such that the method also includes determining a deterioration trend:

[0048] Linear regression analysis is performed on a series of continuous operating health indices of a preset length, including the latest operating health index, to calculate the slope of its changing trend.

[0049] If the slope is negative and its absolute value is greater than the preset trend deterioration judgment threshold, the device health status is determined to be showing an accelerated deterioration trend, and a corresponding alarm is triggered.

[0050] The present invention also provides an adaptive substation equipment condition monitoring system, the system comprising:

[0051] Standardization module: Used to collect multi-source monitoring data from power equipment and perform standardization processing to generate standard time-series data;

[0052] Trajectory construction module: used to divide standard time series data into time windows based on working conditions, extract feature parameters within each window, and construct feature trajectories in chronological order;

[0053] Reference Library Module: Used to build and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions based on historical health data;

[0054] Index generation module: used to compare real-time feature trajectories with standard reference points and health tolerance boundaries of corresponding working conditions in the adaptive reference library, calculate the deviation and generate the operating health index;

[0055] Status determination module: used to determine the dynamic early warning threshold based on historical data of the operating health index, and to determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

[0056] This invention provides an adaptive substation equipment operating condition monitoring system and method. The method comprises: S1: collecting multi-source monitoring data from the substation equipment and standardizing it to generate standard time-series data; S2: dividing the standard time-series data into time windows based on operating conditions, extracting feature parameters within each window, and constructing feature trajectories in chronological order; S3: establishing and updating an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions based on historical health data; S4: comparing the real-time feature trajectories with the standard reference points and health tolerance boundaries corresponding to the operating conditions in the adaptive reference library, calculating the deviation, and generating an operating health index; S5: determining a dynamic early warning threshold based on historical data of the operating health index, and judging the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and a preset rigid alarm threshold. The beneficial effects include:

[0057] 1. By dividing the monitoring data into stable operating condition windows, feature extraction is ensured to be carried out under quasi-static conditions of the equipment, thereby effectively eliminating transient process interference. The extracted electrothermal conversion feature parameters, heat transfer feature parameters, and electromechanical coupling feature parameters directly quantify the equipment's cooling efficiency, insulating oil circuit status, and core mechanical status, respectively, and have clear physical meaning. By constructing feature trajectories in chronological order, the high-dimensional time-series data stream is transformed into a low-dimensional state space path. This path characterizes the evolution process of the internal coordination relationship of the equipment, thereby realizing an intuitive depiction and tracking of the equipment's health deterioration process.

[0058] 2. By classifying operating condition subsets according to load rate intervals and calculating their statistical center and fluctuation range, standard reference points and health tolerance boundaries are formed, thereby establishing personalized health benchmark models for each device under different operating conditions. This method overcomes the misjudgment problem caused by using a unified judgment standard and introduces an incremental update mechanism based on new health data, enabling the reference library to adaptively track the performance changes of the device due to normal aging, avoiding long-term false alarms caused by benchmark drift, and ensuring the applicability and accuracy of the detection system throughout its entire life cycle.

[0059] 3. By calculating the normalized deviation of each feature point relative to its corresponding standard reference point, the absolute numerical difference is transformed into a relative proportion relative to the historical normal fluctuation range. This measurement method is not sensitive to the dimensions and absolute change of the parameters themselves, but focuses on identifying whether there is a deviation from the normal distribution. Furthermore, the average deviation and the deviation fluctuation are combined to generate an operational health index, which not only captures the systematic offset of feature points, but also sensitively reflects the decline in operational stability, thereby enabling a more comprehensive and accurate quantification of the morphological consistency between the entire feature trajectory and the health benchmark.

[0060] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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. In the drawings:

[0062] Figure 1 A flowchart illustrating an adaptive method for detecting the operating conditions of power equipment is shown as an exemplary embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram illustrating the structure of an adaptive power equipment condition monitoring system, which is an exemplary embodiment of the present invention. Detailed Implementation

[0064] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0065] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0066] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0067] Example 1:

[0068] An adaptive method for detecting the operating conditions of power equipment, such as Figure 1 As shown, it includes:

[0069] S1: Collect multi-source monitoring data of power equipment and perform standardization processing to generate standard time series data;

[0070] S2: Based on the time windows of standard time series data divided by working conditions, feature parameters are extracted in each window and feature trajectories are constructed in chronological order;

[0071] S3: Based on historical health data, establish and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions;

[0072] S4: Compare the real-time feature trajectory with the standard reference points and health tolerance boundaries of the corresponding working conditions in the adaptive reference library, calculate the deviation, and generate the operating health index;

[0073] S5: Determine the dynamic early warning threshold based on historical data of the operating health index, and determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

[0074] The present invention is further configured such that S1 includes:

[0075] Using the data stream with the highest sampling frequency among the multi-source monitoring data as the time reference, the remaining data streams are uniformly aligned to the timestamp sequence of the reference through an interpolation algorithm. The invention is further configured such that the multi-source monitoring data includes electrical quantity monitoring data, thermal quantity monitoring data, mechanical quantity monitoring data, and operating condition label data. The electrical quantity monitoring data includes the effective value of the current and active power on the high-voltage side of the transformer; the thermal quantity monitoring data includes the top oil temperature and the winding hot spot temperature; the mechanical quantity monitoring data includes the effective value of the core vibration acceleration; and the operating condition label data includes load rate range labels used to characterize the operating conditions of the equipment.

[0076] A statistical discrimination method based on a sliding time window is used to identify and replace transient abnormal data points in each monitoring data sequence;

[0077] Identify and mark the time periods with consecutive missing data in each monitoring data sequence;

[0078] Multi-source monitoring data, after time alignment, noise filtering, and missing data tagging, are combined according to a preset dimensional order to generate a standard time-series data vector. Specifically, electrical quantity monitoring data originates from protection and control devices connected to the transformer's electrical circuit, acquiring RMS current and active power at a first preset sampling frequency; thermal quantity monitoring data is obtained through digital temperature sensors deployed on the transformer body, acquiring top oil temperature and winding hot spot temperature at a second preset sampling frequency lower than the first preset sampling frequency; mechanical quantity monitoring data is acquired through accelerometers installed at designated locations on the transformer tank wall, which continuously acquires vibration signals at a third preset sampling frequency higher than the first and second preset sampling frequencies, and is integrated with... The local processing unit calculates the effective value of its vibration acceleration in real time; the load rate interval label data characterizing the equipment's operating condition is periodically calculated by the upper-level monitoring system based on real-time active power, and divided into several discrete operating condition levels according to preset load level thresholds. For example, the load rate interval of 0% to 30% is defined as the first load level, the interval of 30% to 60% as the second load level, and the interval of 60% to 100% as the third load level. To solve the asynchronous problem caused by the inconsistency between sampling frequency and timestamp of multi-source sensor data, the mechanical quantity monitoring data stream with the highest sampling frequency is used as the reference time series; for other data streams with high sampling frequencies, such as electrical quantity monitoring data, an interpolation algorithm is used. Based on the original sampling point values ​​and timestamps, the corresponding values ​​at each sampling moment in the baseline time series are calculated. For thermal quantity monitoring data streams with low sampling frequencies and non-continuously updated operating condition label data, a forward hold method is used. That is, when a new data point or label value is obtained, its value is held and assigned to all baseline time points from that point until the next new value arrives. In this way, all monitoring data are unified to the same high-precision time base in the time dimension, achieving strict data synchronization. After completing the spatiotemporal alignment of multi-source monitoring data, a data quality cleaning step is performed. The core purpose of this step is to filter out noise and interference that are not related to equipment status in the original signal and to identify invalid data intervals to improve the subsequent analysis. Reliability: This processing method is mainly aimed at electrical quantity monitoring data that are susceptible to transient power grid disturbances. Specifically, a sliding time window covering a preset time length is used to scan and analyze the unified time series point by point. For the current data point at the center of the window, the arithmetic mean and standard deviation of all data points in the sliding window are calculated. If the absolute deviation between the value of the current data point and the arithmetic mean exceeds a preset multiple threshold of the standard deviation, the point is determined to be transient impulse noise, i.e., transient abnormal data point. For the data points determined to be abnormal, their values ​​are replaced with the median value of all data points in the same sliding window. In this way, transient interference is effectively suppressed while maintaining the overall trend and continuity of the data series to the greatest extent.Furthermore, for all types of monitoring data sequences, continuous missing data identification is performed to detect invalid data intervals of a preset time length caused by communication link interruptions, temporary sensor failures, or signals exceeding the measurement range. When multiple consecutive data points in a data stream are invalid on a unified time series, that period will be identified and marked as continuously missing, ensuring that unreliable data periods can be reasonably avoided or labeled when calculating subsequent feature parameters, thereby preventing misjudgments caused by data quality issues. After time alignment and quality cleaning, the values ​​of each monitoring parameter at each reference time point are combined according to a preset fixed parameter dimension order to generate an independent standard time series data vector. The elements of this vector are, in order, the effective value of current, active power, top oil temperature, winding hot spot temperature, and effective value of vibration acceleration. Simultaneously, a load rate interval label vector synchronized with the time axis is generated, where each label value corresponds one-to-one with the equipment operating condition at the corresponding time.

[0079] The present invention is further configured such that S2 includes:

[0080] Based on the operating condition label data, periods of continuous load rate ranges with a duration greater than the preset stable threshold are divided into stable operating condition windows.

[0081] Within each stable operating window, based on standard time-series data, the electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters are calculated.

[0082] The electrothermal conversion characteristic parameters are obtained by calculating the slope of the linear regression between the active power and the rate of change of the winding hot spot temperature within the window.

[0083] The heat transfer characteristic parameters are obtained by calculating the time delay corresponding to the maximum value of the cross-correlation function between the winding hot spot temperature and the top oil temperature within the window.

[0084] The electromechanical coupling characteristic parameters are obtained by fitting a quadratic function relationship between the effective value of vibration acceleration and the effective value of current within the window, and by calculating the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient.

[0085] The electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters calculated for each steady-state operating window are combined to form a characteristic point.

[0086] The feature points corresponding to each steady-state operating condition window are arranged in chronological order to construct a feature trajectory. Specifically, the load rate interval label vector is traversed to identify the period when the equipment is continuously in a stable operating state. The identification condition is that the continuous duration for which the load rate interval label value remains unchanged must be greater than a preset load rate stability threshold. This threshold is set according to the thermal dynamic characteristics of the equipment to ensure that the thermal state of the equipment is in quasi-equilibrium condition within the defined period, thereby providing a stable operating condition premise for the subsequent calculation of feature parameters. Each continuous period that meets the duration condition is defined as a steady-state operating condition window, and the start and end times of the window are recorded. Within each steady-state operating condition window... Based on the standard time-series data vector corresponding to the window and using data validity markers, three core feature parameters are calculated. These feature parameters include electrothermal conversion feature parameters, heat transfer feature parameters, and electromechanical coupling feature parameters. The electrothermal conversion feature parameter is used to quantify the winding temperature rise rate caused by a unit power input. Its calculation process is as follows: numerical difference is performed on the winding hot spot temperature data sequence to obtain the temperature rise rate sequence; then, linear regression analysis is performed with the active power data sequence as the independent variable and the temperature rise rate sequence as the dependent variable, and the obtained regression slope is defined as the electrothermal conversion feature parameter. A decrease in this parameter indicates a decrease in winding cooling efficiency. The heat transfer feature parameter is used to quantify the heat transfer from the winding... The time delay required for heat transfer to the top oil layer is calculated as follows: Cross-correlation analysis is performed on the winding hot spot temperature data sequence and the top oil temperature data sequence. The time shift that maximizes the cross-correlation coefficient is identified and defined as a heat transfer characteristic parameter. An increase in this parameter reflects an increase in oil circuit circulation resistance or a decrease in oil heat transfer performance. The electromechanical coupling characteristic parameter is used to quantify the nonlinear saturation characteristics of the core vibration response as a function of current. Its calculation process involves: using the current effective value data sequence as input and the vibration acceleration effective value data sequence as output, fitting a quadratic polynomial function. The ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient in the fitted function is calculated and defined as... The electromechanical coupling characteristic parameter is an abnormally large value, indicating that the core or clamping structure may be loose. Based on the calculated electrothermal conversion characteristic parameter, heat transfer characteristic parameter, and electromechanical coupling characteristic parameter, feature points corresponding to the current stable operating condition window are constructed. These feature points are represented as a multi-dimensional vector, and the values ​​of each dimension are determined in sequence by the calculation results of the above three characteristic parameters. After the feature points of all stable operating condition windows are constructed, the corresponding feature points are arranged and connected according to the order of the start time points of each window, thereby forming an ordered discrete point sequence in the multi-dimensional state space defined by the three characteristic parameters, that is, constructing a feature trajectory that represents the evolution path of the equipment's operating state.

[0087] The present invention is further configured such that S3 includes:

[0088] Based on the load rate interval corresponding to the generation of each feature point in the feature trajectory, the feature points corresponding to the historical health data are classified into different load rate intervals to form multiple operating condition subsets.

[0089] For each subset of working conditions, calculate the mean vector and standard deviation vector of all feature points in each feature parameter dimension, determine the mean vector as the standard reference point under that working condition, and determine the health tolerance boundary under that working condition based on the mean vector and standard deviation vector.

[0090] Establish a mapping relationship between load rate range, standard reference point and health tolerance boundary, and store it in the form of structured data to complete the construction of the adaptive reference library;

[0091] During real-time monitoring, new feature points identified as representing the health status of equipment through independent state assessments are assigned to corresponding operating condition subsets based on their load rate ranges. The standard reference point and health tolerance boundary under these operating conditions are then recalculated based on the updated operating condition subsets to update the adaptive reference library. Specifically, all feature trajectories generated by the substation equipment during its historical healthy operation phases are acquired. These historical healthy operation phases include the initial commissioning period or the stable operation cycle after a major overhaul. For each feature point in the feature trajectory, it is categorized into the corresponding load rate range based on the load rate range label recorded in the stable operating condition window where it was generated. The load rate range labels include low load, medium load, and high load. Feature points with the same label constitute a working condition. Subset of operating conditions; through the above operations, the normal operating modes of the equipment under different load levels are separated, laying the foundation for establishing a multi-condition health benchmark; for each condition subset, the following steps are performed to generate a standard reference point and health tolerance boundary: First, calculate the average value of all feature points in the subset across each feature parameter dimension to obtain a mean vector, and define this mean vector as the standard reference point under the load rate interval, used to characterize the statistical center position of the health status under this condition in the feature space; then, calculate the standard deviation of all feature points in the subset across each feature parameter dimension to obtain a standard deviation vector; based on the standard deviation vector, extend the standard reference point as the center in each dimension in both positive and negative directions by a preset multiple of the standard deviation, thereby forming a... A closed numerical interval in the feature space is defined as the health tolerance boundary of the corresponding load rate interval, used to quantify the normal fluctuation range of characteristic values ​​allowed for the health status under the operating condition. The calculated standard reference points and health tolerance boundaries of each load rate interval are structured and stored to construct an adaptive reference library. An independent record is created for each load rate subset that has completed statistical analysis, each record containing a load rate interval identifier, a corresponding standard reference point vector, and a corresponding health tolerance boundary vector. By establishing and associating the mapping relationship between each load rate interval and its corresponding health benchmark model, an adaptive reference library integrating multi-condition health status mapping relationships is constructed. The health benchmark model includes... A complete record of standard reference points and health tolerance boundaries is required. After real-time monitoring of the equipment is implemented, an online dynamic update mechanism is necessary to maintain the long-term effectiveness of the adaptive reference library. When newly generated feature points are confirmed as representing the health status of the equipment through independent verification methods such as routine power outage tests and oil chromatography analysis, an update process is triggered: First, based on the load rate interval label of the new feature point, it is assigned to the corresponding operating condition subset in the adaptive reference library. Subsequently, based on the updated complete feature point set of this subset, the statistical calculation of standard reference points and health tolerance boundaries is re-executed, and the corresponding load rate interval record in the library is updated accordingly. This update process is constrained by preset conditions, such as only being executed when the time since the last update exceeds a preset period or when new data accumulates to a preset quantity.The adaptive reference library also includes an elimination mechanism. If a certain load rate range is not matched by real-time data or its corresponding operating condition does not reappear within a preset time period, the health baseline model for that range will be marked as dormant to optimize the allocation of real-time computing resources.

[0092] The present invention is further configured such that S4 includes:

[0093] Based on the load rate range corresponding to each feature point in the real-time feature trajectory, the corresponding standard reference point and health tolerance boundary are matched and obtained from the adaptive reference library;

[0094] Based on the numerical deviation of each feature point from its standard reference point in each feature parameter dimension, and the threshold interval width of the health tolerance boundary in each feature parameter dimension, the normalized deviation of each feature point is calculated. The invention further specifies that calculating the normalized deviation of each feature point specifically includes: calculating the difference between the value of the feature point in each feature parameter dimension and the value of its standard reference point in the corresponding dimension; dividing the difference by the threshold interval width of the health tolerance boundary in the corresponding feature parameter dimension to obtain the normalized component of that dimension; summing the squares of the normalized components of all feature parameter dimensions, and performing a square root operation on the summed result to obtain the normalized deviation of the feature point.

[0095] Calculate the arithmetic mean of the normalized deviations of all feature points in the real-time feature trajectory, and use it as the average deviation of the trajectory.

[0096] Calculate the standard deviation of the normalized deviation of all feature points relative to the average deviation, and use it as the deviation fluctuation of the trajectory.

[0097] The average deviation is added to the deviation fluctuation weighted by a preset weighting coefficient, and this sum is mapped using a monotonically decreasing function to generate an operational health index. Specifically, for a real-time feature trajectory composed of ordered feature points within a preset continuous time period, the corresponding record in the adaptive reference library is indexed and matched based on the load rate interval label of each feature point to extract the standard reference point and health tolerance boundary of that load rate interval. The normalized deviation of each feature point is calculated, which is used to quantify the degree of deviation of the feature point relative to its corresponding historical health benchmark and to eliminate differences in the units and normal fluctuation range of different feature parameters. The calculation process for deviation is as follows: For each feature parameter dimension, calculate the difference between the feature point value and the corresponding value of the standard reference point; divide this difference by the threshold interval width of the health tolerance boundary of that feature parameter dimension to obtain the normalized component of that dimension, and iterate through all feature parameter dimensions to obtain their normalized components respectively; then, square the normalized components of each feature parameter dimension, sum the squared values, and perform a square root operation on the sum to obtain the normalized deviation of that feature point. This normalized deviation is used to characterize the degree to which the equipment status deviates from the historical health benchmark in the comprehensive feature parameter dimensions. The larger the normalized deviation value, the greater the deviation of the equipment. The more significant the deviation of the equipment status from the historical health benchmark in terms of comprehensive feature dimensions, the more statistically significant the deviation. Based on the normalized deviation of all feature points, a statistic is calculated to characterize the overall consistency of the real-time feature trajectory. First, the arithmetic mean of the normalized deviation of all feature points in the real-time feature trajectory is calculated as the average deviation, used to quantify the systematic deviation of the equipment status relative to the historical health benchmark during this period. Second, the standard deviation of the normalized deviation of all feature points relative to the average deviation is calculated as the deviation fluctuation, used to reflect the dispersion and stability of the equipment operating status during this period. The average deviation is then compared with a pre-weighted... The deviation fluctuations weighted by coefficients are summed to obtain a comprehensive deviation index. This comprehensive deviation index is then mapped to an operational health index through a preset monotonically decreasing function. The monotonically decreasing function is configured such that when the comprehensive deviation index is zero, the output value is the preset maximum value, and as the comprehensive deviation index increases, the output value monotonically decreases and approaches zero. The operational health index is thus mapped to a predefined numerical range. The operational health index quantifies the degree of deviation of the equipment's operating trajectory from the historical health benchmark in terms of both overall deviation and fluctuation stability. The higher the operational health index value, the healthier and more stable the equipment's operating status.

[0098] The present invention is further configured such that S5 includes:

[0099] Based on a sliding window of a preset time length, moving statistics are performed on the historical operational health index series to calculate the moving mean and moving standard deviation of the historical operational health index within the window.

[0100] The dynamic warning threshold is obtained by subtracting the moving standard deviation from the moving mean and multiplying it by a preset sensitivity coefficient.

[0101] The latest operational health index is compared with both the dynamic early warning threshold and the preset rigid alarm threshold, and the equipment status is determined based on the comparison results.

[0102] When the operating health index is greater than or equal to the dynamic early warning threshold, the equipment is determined to be in normal condition;

[0103] When the operating health index is less than the dynamic early warning threshold and the operating health index is greater than or equal to the preset rigid alarm threshold, the device is determined to be in an early warning state.

[0104] When the operating health index is less than a preset rigid alarm threshold, the equipment is determined to be in an alarm state. Specifically, firstly, a moving statistical analysis is performed on the operating health index sequence based on a sliding window of a preset time length to calculate the dynamic early warning threshold: a historical operating health index sequence generated at preset time intervals is maintained, and a sliding statistical window covering a preset continuous time period is set, such as including data points from the past week; in each calculation cycle, all operating health indices within the window are extracted, and their moving mean and moving standard deviation are calculated; the moving mean represents the statistical center level of the recent operating health status of the equipment, and the moving standard deviation quantifies the normal fluctuation range of the health status; the dynamic early warning threshold is obtained by subtracting the product of the moving standard deviation and a preset sensitivity coefficient from the moving mean. This sensitivity coefficient is an adjustable parameter used to control the system's sensitivity to abnormal fluctuations; the above method enables the dynamic early warning threshold to adaptively track the health benchmark drift caused by gradual changes in normal performance or slow changes in the environment; secondly, a preset... A fixed rigid alarm threshold is used, and the device status is determined based on this rigid alarm threshold and the dynamic early warning threshold obtained in the previous steps. The rigid alarm threshold is a fixed lower limit, independent of the dynamic early warning threshold, set based on historical data analysis and engineering experience, and is used to characterize the critical point of severe deterioration in health status. The device status determination logic is as follows: The latest operating health index is obtained and compared with both the dynamic early warning threshold and the rigid alarm threshold. If the operating health index is greater than or equal to the dynamic early warning threshold, the device is determined to be in a normal state. If the operating health index is less than the dynamic early warning threshold but greater than or equal to the preset rigid alarm threshold, the device is determined to be in an early warning state, indicating that the health level has deviated from the recent normal fluctuation range. At this time, an early warning record is generated, and the difference between the early warning index and the dynamic early warning threshold is recorded. If the operating health index is less than the preset rigid alarm threshold, the device is determined to be in an alarm state, indicating that the health status has deteriorated to the point requiring immediate intervention, and a high-level alarm notification is triggered.

[0105] The present invention is further configured such that the method also includes determining a deterioration trend:

[0106] Linear regression analysis is performed on a series of continuous operating health indices of a preset length, including the latest operating health index, to calculate the slope of its changing trend.

[0107] If the slope is negative and its absolute value is greater than the preset trend deterioration judgment threshold, the equipment health status is determined to be showing an accelerated deterioration trend, and a corresponding alarm is triggered. Specifically, by analyzing the continuous change trend of the operating health index, potential and continuous state deterioration processes are identified to achieve earlier warnings compared to judgments based on instantaneous thresholds. The process is as follows: extract a continuous historical operating health index sequence of a preset time length, including the latest operating health index; perform linear regression analysis based on this ordered numerical sequence to fit a straight line characterizing the change trend of the operating health index within the time period, and calculate the slope of the line, which quantifies the average rate of change of the operating health index within the time period; then, compare the calculated slope with the preset trend deterioration judgment threshold. If the slope is negative and its absolute value is greater than the trend deterioration judgment threshold, the equipment health status is determined to show a clear accelerated deterioration trend. If this condition is met, an independent trend deterioration alarm will be triggered regardless of whether the current operating health index is below the dynamic warning threshold. This deterioration trend judgment mechanism enhances the early detection capability of gradual and latent faults.

[0108] Example 2:

[0109] Please see Figure 2 An exemplary adaptive substation equipment condition monitoring system includes:

[0110] Standardization module: Used to collect multi-source monitoring data from power equipment and perform standardization processing to generate standard time-series data;

[0111] Trajectory construction module: used to divide standard time series data into time windows based on working conditions, extract feature parameters within each window, and construct feature trajectories in chronological order;

[0112] Reference Library Module: Used to build and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions based on historical health data;

[0113] Index generation module: used to compare real-time feature trajectories with standard reference points and health tolerance boundaries of corresponding working conditions in the adaptive reference library, calculate the deviation and generate the operating health index;

[0114] Status determination module: used to determine the dynamic early warning threshold based on historical data of the operating health index, and to determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

[0115] It should be noted that the adaptive substation equipment condition detection system and the adaptive substation equipment condition detection method provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the adaptive substation equipment condition detection system provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0116] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive method for detecting the operating conditions of power equipment, characterized in that, include: S1: Collect multi-source monitoring data of power equipment and perform standardization processing to generate standard time series data; S2: Based on the time windows of standard time series data divided by working conditions, feature parameters are extracted in each window and feature trajectories are constructed in chronological order; S3: Based on historical health data, establish and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions; S4: Compare the real-time feature trajectory with the standard reference points and health tolerance boundaries of the corresponding working conditions in the adaptive reference library, calculate the deviation, and generate the operating health index; S5: Determine the dynamic early warning threshold based on historical data of the operating health index, and determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

2. The adaptive substation equipment operating condition detection method according to claim 1, characterized in that, S1 includes: Using the data stream with the highest sampling frequency among the multi-source monitoring data as the time reference, the remaining data streams are uniformly aligned to the timestamp sequence of the reference through an interpolation algorithm; A statistical discrimination method based on a sliding time window is used to identify and replace transient abnormal data points in each monitoring data sequence; Identify and mark the time periods with consecutive missing data in each monitoring data sequence; The multi-source monitoring data, after time alignment, noise filtering, and missing data labeling, are combined according to a preset dimensional order to generate a standard time-series data vector.

3. The adaptive substation equipment operating condition detection method according to claim 2, characterized in that, The multi-source monitoring data includes electrical quantity monitoring data, thermal quantity monitoring data, mechanical quantity monitoring data, and operating condition label data. The electrical quantity monitoring data includes the effective value of the current and active power on the high-voltage side of the transformer. The thermal quantity monitoring data includes the top oil temperature and the winding hot spot temperature. The mechanical quantity monitoring data includes the effective value of the core vibration acceleration. The operating condition label data includes load rate range labels used to characterize the operating conditions of the equipment.

4. The adaptive substation equipment operating condition detection method according to claim 1, characterized in that, S2 includes: Based on the operating condition label data, periods of continuous load rate ranges with a duration greater than the preset stable threshold are divided into stable operating condition windows. Within each stable operating window, based on standard time-series data, the electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters are calculated. The electrothermal conversion characteristic parameters are obtained by calculating the slope of the linear regression between the active power and the rate of change of the winding hot spot temperature within the window. The heat transfer characteristic parameters are obtained by calculating the time delay corresponding to the maximum value of the cross-correlation function between the winding hot spot temperature and the top oil temperature within the window. The electromechanical coupling characteristic parameters are obtained by fitting a quadratic function relationship between the effective value of vibration acceleration and the effective value of current within the window, and by calculating the ratio of the absolute value of the quadratic term coefficient to the absolute value of the linear term coefficient. The electrothermal conversion characteristic parameters, heat transfer characteristic parameters, and electromechanical coupling characteristic parameters calculated for each steady-state operating window are combined to form a characteristic point; The feature points corresponding to each steady-state operating condition window are arranged in chronological order to construct the feature trajectory.

5. The adaptive substation equipment operating condition detection method according to claim 1, characterized in that, S3 includes: Based on the load rate interval corresponding to the generation of each feature point in the feature trajectory, the feature points corresponding to the historical health data are classified into different load rate intervals to form multiple operating condition subsets. For each subset of working conditions, calculate the mean vector and standard deviation vector of all feature points in each feature parameter dimension, determine the mean vector as the standard reference point under that working condition, and determine the health tolerance boundary under that working condition based on the mean vector and standard deviation vector. Establish a mapping relationship between load rate range, standard reference point and health tolerance boundary, and store it in the form of structured data to complete the construction of adaptive reference library; During real-time monitoring, new feature points that are identified as characterizing the health status of equipment through independent status assessment are assigned to the corresponding operating condition subset according to their load rate range. The standard reference point and health tolerance boundary under the updated operating condition subset are then recalculated to complete the update of the adaptive reference library.

6. The adaptive substation equipment operating condition detection method according to claim 1, characterized in that, S4 includes: Based on the load rate range corresponding to each feature point in the real-time feature trajectory, the corresponding standard reference point and health tolerance boundary are matched and obtained from the adaptive reference library; Based on the numerical deviation of each feature point from its standard reference point in each feature parameter dimension, and the threshold interval width of the health tolerance boundary in each feature parameter dimension, the normalized deviation of each feature point is calculated. Calculate the arithmetic mean of the normalized deviations of all feature points in the real-time feature trajectory, and use it as the average deviation of the trajectory. Calculate the standard deviation of the normalized deviation of all feature points relative to the average deviation, and use it as the deviation fluctuation of the trajectory. The average deviation is added to the deviation fluctuation weighted by a preset weighting coefficient, and the sum is mapped by a monotonically decreasing function to generate the operational health index.

7. The adaptive substation equipment operating condition detection method according to claim 6, characterized in that, The calculation of the normalized deviation of each feature point specifically includes: Calculate the difference between the value of the feature point in each dimension of the feature parameter and the value of its standard reference point in the corresponding dimension; Divide the difference by the threshold interval width of the health tolerance boundary of the corresponding feature parameter dimension to obtain the normalized component of that dimension. The normalized deviation of a feature point is obtained by summing the squares of the normalized components of all feature parameter dimensions and then taking the square root of the sum.

8. The adaptive substation equipment operating condition detection method according to claim 1, characterized in that, S5 includes: Based on a sliding window of a preset time length, moving statistics are performed on the historical operational health index series to calculate the moving mean and moving standard deviation of the historical operational health index within the window. The dynamic warning threshold is obtained by subtracting the moving standard deviation from the moving mean and multiplying it by a preset sensitivity coefficient. The latest operational health index is compared with both the dynamic early warning threshold and the preset rigid alarm threshold, and the equipment status is determined based on the comparison results. When the operating health index is greater than or equal to the dynamic early warning threshold, the equipment is determined to be in normal condition; When the operating health index is less than the dynamic early warning threshold and the operating health index is greater than or equal to the preset rigid alarm threshold, the device is determined to be in an early warning state. When the operating health index is lower than the preset rigid alarm threshold, the device is determined to be in an alarm state.

9. The adaptive substation equipment operating condition detection method according to claim 8, characterized in that, The method also includes determining the deterioration trend: Linear regression analysis is performed on a series of continuous operating health indices of a preset length, including the latest operating health index, to calculate the slope of its changing trend. If the slope is negative and its absolute value is greater than the preset trend deterioration judgment threshold, the device health status is determined to be showing an accelerated deterioration trend, and a corresponding alarm is triggered.

10. An adaptive substation equipment condition detection system, used to implement the adaptive substation equipment condition detection method according to any one of claims 1-9, characterized in that, include: Standardization module: Used to collect multi-source monitoring data from power equipment and perform standardization processing to generate standard time-series data; Trajectory construction module: used to divide standard time series data into time windows based on working conditions, extract feature parameters within each window, and construct feature trajectories in chronological order; Reference Library Module: Used to build and update an adaptive reference library containing standard reference points and health tolerance boundaries under different operating conditions based on historical health data; Index generation module: used to compare real-time feature trajectories with standard reference points and health tolerance boundaries of corresponding working conditions in the adaptive reference library, calculate the deviation and generate the operating health index; Status determination module: used to determine the dynamic early warning threshold based on historical data of the operating health index, and to determine the equipment status by comparing the real-time operating health index with the dynamic early warning threshold and the preset rigid alarm threshold.

Images