Operation state analysis method and system for transformer

Abstract

The invention relates to the technical field of electric data processing, in particular to an operation state analysis method and system for a transformer, and the method comprises the steps: obtaining the operation parameters of the transformer; according to the size of the load rate of the transformer, power consumption time periods are divided every day, for any power consumption time period, a health sample index of data in the power consumption time period is calculated, health sample data are screened out, and a health data set is constructed. According to the method, the health assessment model of each time period is established through the mahalanobis distance, and weighting processing is performed in combination with the anomaly index, so that the model has adaptive adjustment capability under different health levels, false alarm can be avoided, sensitivity to real anomaly can be improved, then the anomaly threshold is determined by using chi-square distribution, and the health assessment accuracy is improved. The accuracy of monitoring the operation state of the transformer is improved, and the false alarm rate and the missing alarm rate are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical data processing, and more particularly to a method and system for analyzing the operating status of a transformer. Background Art

[0002] As the core hub of power systems, transformers' safe and stable operation is directly related to the reliability of the entire power grid. Therefore, collecting multi-dimensional operating parameters such as transformer current, voltage, and oil temperature, and using intelligent algorithms to conduct real-time, accurate analysis and early warning of their operating status, has become a key technology for ensuring power grid security.

[0003] Current transformer condition monitoring methods typically rely on placing sensors such as current, voltage, and temperature at key locations to collect operating parameters. These sensors then determine whether the operating status is abnormal by setting thresholds or constructing statistical models. For example, some methods identify anomalies based on the fluctuation amplitude of a single parameter, while others utilize multidimensional data to construct Mahalanobis distance models or other statistical learning models to detect anomalies in newly collected data. While these methods have improved the ability to identify abnormal transformer conditions to some extent, they still have some limitations.

[0004] However, most condition assessment models require training based on pure, healthy data. In real-world scenarios, historical operating data is often unlabeled and contains a mix of normal, sub-healthy, and even early-stage fault conditions, making it difficult to directly use to construct an accurate assessment benchmark. Secondly, the operating conditions of a transformer are closely related to its load level. During peak, steady, and off-peak periods, the normal range and pattern of parameter fluctuations vary significantly. Using a single, static assessment model is difficult to adapt to these dynamic changes and is prone to high false positives and false negatives. Summary of the Invention

[0005] The present invention provides a method and system for analyzing the operating status of a transformer, aiming to solve the problem in related technologies that the operating conditions of a transformer are closely related to its load level, and that the normal parameter fluctuation ranges and patterns vary significantly during different power consumption periods, such as peak, stable, and valley periods. The use of a single, static evaluation model is difficult to adapt to such dynamic changes and is prone to high false alarm or missed alarm rates.

[0006] In a first aspect, the present invention provides a method for analyzing the operating status of a transformer, comprising: obtaining operating parameters of the transformer; dividing daily power consumption periods according to the load rate of the transformer; calculating a health sample index for the data within each power consumption period, screening out healthy sample data, and constructing a health dataset, wherein the health sample index reflects the volatility of the data within the power consumption period and the similarity between the power consumption period and its historical cycles; constructing a Mahalanobis distance health assessment model corresponding to the power consumption period based on the health dataset; obtaining new operating parameters, and calculating the Mahalanobis distance value of the new operating parameters using the Mahalanobis distance health assessment model corresponding to the power consumption period to which they belong, and performing transformer status analysis based on the Mahalanobis distance value. Dynamically dividing the operating time into periods according to the transformer load rate and constructing a dedicated assessment model for each period solves the problem in existing technologies that a single static model is difficult to adapt to the large differences in the normal fluctuation range of transformer parameters under different operating conditions such as peak, stable, and low peak conditions, significantly improving the adaptability and accuracy of status analysis. In addition, based on the health sample index of data volatility and historical cycle similarity, it is possible to screen out samples that best represent the health status of the period from unlabeled historical data mixed with various states to build an evaluation model, making the model benchmark more reliable, improving the accuracy and automation level of transformer operating status analysis, and effectively reducing false alarm and missed alarm rates.

[0007] Furthermore, the daily electricity consumption period is divided, including: using a sliding window to calculate the average load rate of all times in the window; if the average load rate is greater than a first threshold, the corresponding period is divided into a peak period; if the average load rate is greater than or equal to a second threshold and less than or equal to the first threshold, the corresponding period is divided into a stable period; if the average load rate is less than the second threshold, the corresponding period is divided into a low-peak period.

[0008] Furthermore, calculating the health sample index of the data within the power usage period involves setting a window for the operating parameters of the power usage period, sliding the window along the operating parameters, and calculating the health sample index of the operating parameters within the window after each slide. By setting a sliding window to continuously calculate the health sample index, the health characteristics of each tiny segment in the data stream can be more precisely captured, achieving a comprehensive and continuous assessment of all data segments throughout the entire power usage period.

[0009] Furthermore, healthy sample data is screened out by obtaining a healthy sample index for the operating parameters within the window after each slide, and selecting the operating parameters within the window corresponding to the maximum healthy sample index as the healthy sample data. This method can largely eliminate potential interference from sub-health or minor anomalies in historical data, ensuring the highest purity of the constructed health assessment model, thereby making the model more sensitive to deviations from normal conditions.

[0010] Furthermore, based on the size of the Mahalanobis distance value, the transformer status is analyzed, including: if the Mahalanobis distance value is greater than the abnormal threshold, the operating status of the transformer is determined to be abnormal, and a recording or alarm mechanism is triggered; if the Mahalanobis distance value is less than or equal to the abnormal threshold, the operating status of the transformer is determined to be normal.

[0011] Furthermore, a Mahalanobis distance health assessment model corresponding to the electricity consumption period is constructed, including: obtaining a health data set for the electricity consumption period, calculating the mean vector of each dimension in the health data set and the covariance matrix of the health data set, and constructing a Mahalanobis distance health assessment model based on the mean vector and the health data set.

[0012] Furthermore, the system calculates an anomaly index for each power usage period and uses it to weight the Mahalanobis distance value calculated by the evaluation model for that period. This dynamically adjusts the strictness of the judgment based on the overall health of the period, making the model more sensitive during periods of overall stability and more tolerant during periods of overall fluctuation. This enables more intelligent and context-aware anomaly judgment, further reducing false positives.

[0013] Furthermore, the anomaly index for each electricity usage period is calculated. This involves calculating the healthy sample index of the multidimensional data within each window after each sliding window within that period. The ratio of healthy sample indices exceeding a threshold across all windows is then counted as the anomaly index for that period. This method avoids subjective, empirically determined anomaly indexes, providing a reliable basis for calculation. This ensures the objectivity and reproducibility of the weighted correction process, thereby ensuring the stability and reliability of the entire analysis model.

[0014] Furthermore, the method for determining the abnormality threshold includes: determining the abnormality threshold based on chi-square distribution. In a second aspect, the present invention further provides a system for analyzing the operating status of a transformer, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement any one of the above methods for analyzing the operating status of a transformer.

[0015] Beneficial effects (1) By introducing a dynamic load-based power consumption time-segmentation mechanism, transformer operating data can be modeled differently under different operating conditions, avoiding the risk of failure of a single model in full-time monitoring. Furthermore, by constructing a healthy sample index that integrates immediate volatility and historical cyclical similarity, accurate screening of healthy samples is achieved, thereby ensuring the stability of the benchmark data for the evaluation model.

[0016] (2) The Mahalanobis distance is used to establish a health assessment model for each time period, and the abnormality index is combined for weighted processing, so that the model has the ability to adaptively adjust at different health levels, which can not only avoid false alarms but also improve the sensitivity to real abnormalities. Then, the chi-square distribution is used to determine the abnormality threshold, which improves the accuracy of transformer operating status monitoring and effectively reduces the false alarm and missed alarm rates. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. 4 is a flow chart schematically illustrating operation status monitoring of a transformer according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] like Figure 1 As shown, S101: collecting the operating parameters of the transformer and dividing the power consumption time period.

[0020] Specifically, various high-precision sensors deployed at key locations on the transformer collect multi-dimensional operating parameters in real time. These parameters include, but are not limited to, current, voltage, oil temperature, ambient temperature, and winding temperature, and are then combined into a multi-dimensional time series dataset.

[0021] Taking into account that the operating conditions of the transformer are closely related to its load level, and the load level shows obvious periodic changes within a day, this embodiment introduces a dynamic division of power consumption periods based on the load rate. Specifically, for any natural day, the load rate data of the transformer in that natural day is collected in real time, and a sliding window is set for smoothing to remove the influence of noise. As a preferred solution, the size of the sliding window can be set to 1 hour. In other embodiments, the size of the sliding window can also be set to 45 minutes or 30 minutes, which can be adjusted according to the specific implementation scenario. Then, the sliding window is used to slide on the real-time load rate sequence with a step size of 1, and the average value of all load rate samples in the window is calculated. Based on the size of the average value, the 24 hours of a day are dynamically divided into three different power consumption periods, namely the peak period, the stable period and the low-peak period. The specific division process is as follows.

[0022] If the average load rate within the sliding window is greater than the first threshold, the period is divided into a peak period. In this embodiment, the empirical value of the first threshold is 70%. The reason is that this setting is intended to cover the period when industrial production or residential electricity consumption is most concentrated. At this time, the transformer is under the greatest pressure, and the normal range of its operating parameters is significantly different from other periods. If the average load rate within the sliding window is greater than or equal to the second threshold and less than or equal to the first threshold, it is divided into a stable period. This is the normal operating state of the transformer. In this embodiment, the empirical value of the second threshold is 40%. If the average load rate within the sliding window is less than the second threshold, it is divided into a low-peak period. The reason is that the load is the lowest during this period, usually corresponding to nighttime or production breaks, so the average load rate during this period is relatively low. It should be noted that the first threshold is greater than the second threshold.

[0023] Through the above steps, the original continuous data stream can be divided into discrete data segments with clear working condition background, laying the foundation for the subsequent construction of differentiated health assessment models for different working conditions.

[0024] S102: Screening healthy samples at each time period.

[0025] When building an evaluation model for any electricity usage period, we first need to filter out a baseline dataset from the massive amount of historical data that represents a relatively healthy state. Therefore, in this embodiment, we propose a method for constructing a healthy sample index that combines immediate volatility with historical cycle similarity. Taking the off-peak period as an example, the construction process is as follows.

[0026] First, the collected operating parameters are standardized. Then, in this embodiment, a healthy sample index is constructed to quantify whether the multidimensional data in a sliding window of size N during any power usage period is a healthy sample, where N is 20. In other embodiments, N can be 30 or 15, etc. This index is based on the following principles: a truly healthy data segment should not only have minimal internal fluctuations, but also have a fluctuation pattern similar to that of the historical data during the same period.

[0027] Following the above, a method for calculating the health sample index is provided, which is calculated as follows: Where, Indicates the off-peak window The health sample index of multidimensional data, Indicates the During the off-peak period The variance of the dimensional data, Indicates the number of off-peak hours The mean of the variance of the dimensional data, Indicates the number of off-peak hours The maximum value of the variance of the dimensional data, Indicates the number of data dimensions, Indicates the The correction coefficient of multidimensional data in window N during the off-peak period, Represents the exponential function with the natural base e as the base. It is used to measure whether the volatility of a single dimension in a low-peak period deviates from the normal volatility benchmark of the low-peak period, and thus serves as a basic indicator of whether there are healthy samples in the window. The larger the value of , the smaller the health sample index is, divided by It is for normalization.

[0028] This formula not only considers whether the immediate fluctuation amplitude of the multidimensional data within the current window is within the normal range, but also introduces the similarity of the fluctuation pattern with the historical data during the same period as a correction factor, thereby achieving a dual-constraint screening of healthy samples. This approach can effectively avoid misjudgments caused by relying solely on amplitude indicators, making the identification of healthy samples more accurate.

[0029] It should be noted that transformer operating conditions often exhibit regular patterns, such as diurnal cycles and weekday-rest cycles. For example, low-peak periods generally correspond to reduced loads at night, and operating data exhibits relatively stable fluctuation patterns during the same period on different days. If only the instantaneous fluctuation amplitude is considered while ignoring this cyclical pattern, some short-term abnormal fluctuations may be misidentified as normal healthy samples. Therefore, considering cyclical similarity is to leverage the historical regularity of transformer operation, avoid interference from occasional noise in the screening of healthy samples, make the healthy data more representative and stable, and thus improve the reliability and accuracy of subsequent evaluation models.

[0030] For the The correction coefficient of multidimensional data in window N during a low-peak period is obtained as follows: The calculation formula is: Where, It represents the correction coefficient of the multidimensional data in window N during the i-th off-peak period. The correction coefficient reflects the consistency between the data pattern of this window and the overall historical pattern during the same period. Represents the window in the oth dimension of the corresponding period on the mth day The similarity between the corresponding sequence and the sequence corresponding to window N in the o-th dimension during the i-th low-peak period (the similarity is the inverse of the sum of the DTW distance and the preset hyperparameter. The preset hyperparameter avoids the denominator being 0, for example, the preset hyperparameter is 1, and the range of similarity values ​​can be mapped to between 0 and 1). Indicates the number of adjacent days, Indicates the number of data dimensions. In this embodiment, the number of adjacent days is 10 days.

[0031] In the formula, if the window N is highly similar to the fluctuation pattern of the adjacent days in history, then is larger, and the contribution is further amplified after squaring, then If the similarity is generally low, then The introduction of this correction coefficient makes the identification of healthy samples not only rely on whether the fluctuation within the window is normal, but also further combines the similarity of historical periodicity, thereby improving the ability to judge healthy samples.

[0032] In summary, according to the above calculation method, traverse all low-peak periods and find The window corresponding to the maximum value is used, and the multidimensional data in the window is regarded as the healthy data set. The larger the value of , the smoother the fluctuation of the data in the window, and the more consistent its fluctuation pattern is with the historical period. Therefore, the data corresponding to the window can most accurately represent the best health status, thus providing a reliable benchmark sample for subsequent evaluation models.

[0033] S103: Based on the health samples of each electricity consumption period, an evaluation model corresponding to each electricity consumption period is constructed.

[0034] Specifically, taking the off-peak period as an example, the above method is used to obtain the health data set corresponding to the off-peak period. The health data set contains M sample points, and each sample point is a feature vector composed of d dimensional data. Then, the mean vector of each dimension and the covariance matrix of the health data set are calculated. Finally, a Mahalanobis distance health assessment model is constructed based on the mean vector and the health data set. For any new sample, it is input into the Mahalanobis distance health assessment model to obtain the distance value of the new sample in the assessment model. This distance value quantifies the distance of the new sample point from the center of the health data distribution after considering the correlation between dimensions.

[0035] S104: Real-time monitoring of the transformer's operating status based on the evaluation model.

[0036] Once the model is built, it can be used to monitor the transformer's condition in real time. Specifically, when a new data sample is collected, the power consumption period (peak period, stable period, or off-peak period) to which the new data sample belongs is first determined. The Mahalanobis distance assessment model for the corresponding period is then called to calculate its Mahalanobis distance value. During the actual monitoring process, when a new data sample is input into the Mahalanobis distance health assessment model, its Mahalanobis distance value is calculated and compared with a preset anomaly threshold. If the Mahalanobis distance value is less than or equal to the threshold, it indicates that the difference between the new data sample and the healthy data set is within the normal fluctuation range, and the transformer can be determined to be operating normally. Conversely, if the Mahalanobis distance value is greater than the threshold, it indicates that the data sample has significantly deviated from the healthy state and should be determined to be abnormal, triggering a recording or alarm mechanism.

[0037] In another embodiment, an abnormality index for each power consumption period can also be calculated, and the abnormality index of the power consumption period can be used to weight the Mahalanobis distance value calculated by the evaluation model corresponding to the power consumption period. Specifically, obtaining the abnormality index of any power consumption period, taking the off-peak period as an example, includes: determining a partition threshold, the empirical value of the partition threshold is 0.8, then calculating the healthy sample index of the multidimensional data in the window after each window slides in the off-peak period, and counting the proportion of the number of healthy sample indexes in all windows that is greater than the partition threshold to all healthy sample indexes. The larger the proportion, the more stable the fluctuation of most data segments in the period and the healthier the pattern. In other words, the healthier the period as a whole, the calculated Mahalanobis distance value is weighted to make the weighted Mahalanobis distance value smaller. Conversely, the smaller the proportion, the poorer the overall health of the period, with more fluctuations or abnormalities. Once the new sample is significantly different from the healthy data, it is more likely to be a true abnormality. The calculated Mahalanobis distance value is weighted to make the weighted Mahalanobis distance value larger, making the model judgment more stringent.

[0038] The determination of the abnormal threshold includes: in the Mahalanobis distance health assessment model, the degree of difference between the new data sample and the healthy data set can be measured by the Mahalanobis distance. Since the healthy data set follows a multivariate normal distribution, the square value of the Mahalanobis distance follows a chi-square distribution with degrees of freedom equal to the number of data dimensions in statistical theory. Based on this property, the chi-square distribution can be used to determine the demarcation threshold between the healthy state and the abnormal state. Specifically, first determine the degrees of freedom of the chi-square distribution based on the number of characteristic dimensions of the healthy data set; then set a significance level, such as 0.05, 0.01 or 0.001, which reflects the maximum false alarm probability allowed by the system; then find the critical value under the corresponding degrees of freedom and confidence level in the chi-square distribution table. This critical value is the abnormal threshold of the Mahalanobis distance.

[0039] The present invention further provides a system for analyzing the operating status of a transformer. The system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, the method for analyzing the operating status of a transformer according to the first aspect of the present invention is implemented.

[0040] The system further includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and thus will not be described in detail here.

[0041] In the present invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented using computer-readable / executable instructions stored or otherwise retained on such a computer-readable medium.

[0042] The above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.

Claims

1. A method for analyzing the operating status of a transformer, characterized in that: include: Obtaining operating parameters of the transformer; According to the load rate of the transformer, the daily electricity consumption period is divided. For any of the electricity consumption periods, the health sample index of the data in the electricity consumption period is calculated, the healthy sample data is screened out, and a healthy data set is constructed, wherein the health sample index reflects the volatility of the data in the electricity consumption period and the similarity between the electricity consumption period and its historical period; Based on the health data set, constructing a Mahalanobis distance health assessment model corresponding to the electricity consumption period; New operating parameters are obtained, and the Mahalanobis distance health assessment model corresponding to the power consumption period is used to calculate the Mahalanobis distance value of the new operating parameters, and the transformer status is analyzed according to the size of the Mahalanobis distance value.

2. The method for analyzing the operating status of a transformer according to claim 1, characterized in that: The division of daily electricity consumption time includes: Calculating the average load rate at all times within the window using a sliding window; If the average load rate is greater than a first threshold, the corresponding time period is classified as a peak time period; If the average load rate is greater than or equal to the second threshold and less than or equal to the first threshold, the corresponding period is divided into a stable period; If the average load rate is less than the second threshold, the corresponding time period is divided into an off-peak period, wherein the first threshold is greater than the second threshold.

3. The method for analyzing the operating status of a transformer according to claim 1, characterized in that: Calculate the health sample index of the data during the electricity consumption period, including: A window is set for the operating parameters of the electricity consumption period, and the window is made to slide within the operating parameters. After each slide, a health sample index of the operating parameters within the window is calculated.

4. The method for analyzing the operating status of a transformer according to claim 3, characterized in that: Filter out healthy sample data, including: Obtain the healthy sample index of the operating parameters in the window after each sliding, and select the operating parameters in the window corresponding to the maximum healthy sample index as the healthy sample data.

5. The method for analyzing the operating status of a transformer according to claim 1, characterized in that: According to the Mahalanobis distance value, the transformer state analysis is performed, including: If the Mahalanobis distance value is greater than the abnormal threshold, the operating state of the transformer is determined to be abnormal, and a recording or alarm mechanism is triggered; If the Mahalanobis distance value is less than or equal to the abnormality threshold, it is determined that the operating state of the transformer is normal.

6. The method for analyzing the operating status of a transformer according to claim 1, characterized in that: Constructing a Mahalanobis distance health assessment model corresponding to the electricity consumption period, including: A health data set of the electricity usage period is obtained, and the mean vector of each dimension in the health data set and the covariance matrix of the health data set are calculated, and a Mahalanobis distance health assessment model is constructed based on the mean vector and the health data set.

7. The method for analyzing the operating status of a transformer according to claim 1, characterized in that: Also includes: The abnormality index of each power consumption period is calculated, and the abnormality index of the power consumption period is used to perform weighted processing on the Mahalanobis distance value calculated using the evaluation model corresponding to the power consumption period.

8. The method for analyzing the operating status of a transformer according to claim 7, characterized in that: Calculate the abnormal index of each electricity consumption period, including: For any electricity consumption period, calculate the health sample index of the multidimensional data in the window after each sliding of the window in the electricity consumption period; The ratio of the number of healthy sample indices greater than the division threshold in all windows to the total healthy sample indices is counted as the abnormal index of the power consumption period.

9. The method for analyzing the operating status of a transformer according to claim 5, characterized in that: The method for determining the abnormal threshold includes: Determine anomaly thresholds based on the chi-square distribution.

10. A system for analyzing the operating status of a transformer, comprising a processor and a memory, characterized in that: The memory stores a computer program, and the processor executes the computer program to implement the method for analyzing the operating status of a transformer according to any one of claims 1 to 9.

Images