Transformer electrical degree-temperature rise monitoring method based on regression confidence band

Abstract

The invention discloses a transformer electrical degree-temperature rise monitoring method based on a regression confidence band, and relates to the technical field of electrical variable measurement, and the method comprises the steps: collecting the historical operation data of a transformer, calculating an electrical degree variation and a temperature rise value, carrying out the linear regression analysis with the electrical degree variation as an independent variable and the temperature rise value as a dependent variable after preprocessing, and obtaining the electrical degree-temperature rise of the transformer. Obtaining a linear regression equation and a regression standard error, and constructing a dynamic confidence band based on the regression standard error; current data are collected in real time, and if the current temperature rise value exceeds the dynamic confidence band range, it is judged that the transformer operates abnormally and early warning is given out. A traditional fixed threshold value is replaced by the dynamic confidence band, different load working conditions of the transformer can be self-adapted, the false alarm rate and the missing report rate are remarkably reduced, and the accuracy and the reliability of temperature rise monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology, specifically to a transformer power consumption-temperature rise monitoring method based on regression confidence bands. Background Technology

[0002] In existing transformer operation status monitoring technologies, temperature monitoring is a crucial link in ensuring the safe and stable operation of equipment. Traditionally, engineering practice commonly employs a fixed threshold method to monitor transformer temperature rise. This involves pre-setting a fixed upper limit for temperature rise, and when the real-time monitored temperature rise exceeds this threshold, the system determines it as abnormal and issues an alarm. This method is simple to implement and requires minimal computation, and was widely used in early power system monitoring. In addition, some monitoring systems combine single variables such as load rate or current to construct simplified segmented thresholds or static empirical formulas, attempting to coarsely adjust the alarm boundary under different operating conditions. Another type of existing technology uses statistical methods based on historical data, such as calculating the mean and standard deviation of temperature rise within a specific load range, and using "mean ± k times standard deviation" as the alarm boundary. This type of method, to some extent, considers the distribution characteristics of historical operating data. However, the aforementioned existing technologies all have significant limitations: the fixed threshold method cannot adapt to the dynamic changes in transformer load, and is prone to false alarms due to naturally lower temperature rise under low load conditions, while it may miss alarms due to normal temperature rise under high load conditions; although the static empirical formula based on segmented thresholds has some improvement, its boundary setting still relies on manual experience and is difficult to accurately reflect the actual operating characteristics of the equipment; while the method based on simple statistical intervals utilizes historical data, it usually calculates the boundary independently for discrete load intervals, without establishing a continuous physical relationship model, resulting in boundary jumps and an inability to characterize the intrinsic linear relationship between electricity change and temperature rise. Overall, existing technologies lack a temperature rise anomaly judgment mechanism that can dynamically adapt to different load conditions, establish a continuous functional relationship, and possess statistical confidence significance, leading to insufficient adaptive capability of the monitoring system and difficulty in balancing false alarm and missed alarm rates. Summary of the Invention

[0003] To dynamically adapt to different load conditions and establish a temperature rise anomaly judgment mechanism with a continuous functional relationship and statistical confidence significance, this invention proposes a transformer power consumption-temperature rise monitoring method based on regression confidence band, including the following steps: S1: Collect historical operating data of the transformer, which includes at least ambient temperature, equipment temperature and power readings; S2: Based on historical operating data, calculate the change in electricity consumption and the temperature rise value within a preset time window at each sampling moment, wherein the temperature rise value is the difference between the equipment temperature and the ambient temperature; S3: Preprocess historical operating data, including filtering invalid or abnormal data and truncating the temperature rise value corresponding to the same change in electricity to retain data with a preset intermediate ratio. S4: Using the change in electricity consumption in the preprocessed data as the independent variable and the temperature rise as the dependent variable, perform linear regression analysis to obtain the linear regression equation and the regression standard error; S5: Calculate the dynamic confidence band based on the regression standard error, wherein the upper and lower limits of the dynamic confidence band are the predicted value of the linear regression equation plus and minus the regression standard error by a preset multiple, respectively. S6: Real-time acquisition of the transformer's current operating data, calculation of current electricity consumption change and current temperature rise; S7: Substitute the current change in electricity into the linear regression equation to obtain the predicted temperature rise value, and determine whether the current temperature rise value is within the dynamic confidence band. If yes, return to step S6; otherwise, determine that the transformer is operating abnormally and issue a warning message.

[0004] This invention replaces the traditional fixed threshold with a dynamic confidence band, which can adapt to different load conditions of transformers, significantly reduce false alarm rate and false alarm rate, and improve the accuracy and reliability of temperature rise monitoring.

[0005] Furthermore, in step S2, the preset time window is 1 hour, the change in electricity consumption is the change in electricity consumption reading within 1 hour, and the change in electricity consumption is rounded to an integer.

[0006] Furthermore, in step S3, the temperature rise value corresponding to the same change in electrical energy is truncated to retain data at a preset intermediate ratio. Specifically: For multiple temperature rise values ​​with the same change in electrical energy, sort them by numerical value, remove the data at both ends with a preset ratio, and retain the data in the middle as the effective temperature rise data of the change in electrical energy.

[0007] Furthermore, in step S3, the preprocessing also includes: Historical operating data with a temperature rise greater than or equal to 50°C are filtered out to exclude invalid samples from sensor malfunctions or extreme abnormal operating conditions.

[0008] Furthermore, in step S4, linear regression analysis is performed when the amount of preprocessed data is greater than or equal to a preset amount; otherwise, the regression analysis is skipped and the previously successfully constructed linear regression equation and dynamic confidence band are used.

[0009] Furthermore, in step S5, the preset multiple is 2.0, and the corresponding dynamic confidence band is the linear regression predicted value ± 2 times the regression standard error, and the dynamic confidence band corresponds to a 95% confidence level.

[0010] Furthermore, it also includes the following steps: S8: Store the regression parameters, regression standard error, preset multiples of dynamic confidence band, and goodness of fit of the linear regression equation into the database for use by the external monitoring system for abnormal early warning.

[0011] Furthermore, it also includes backup early warning methods: In step S1, the historical maximum and minimum temperature rise values ​​corresponding to each integer change in electricity are pre-calculated and stored; In step S4, when linear regression analysis cannot be performed or the dynamic confidence band is unavailable, the historical extreme value of temperature rise corresponding to the current change in electricity consumption is used as the alarm boundary.

[0012] Compared with the prior art, the present invention has at least the following beneficial effects: (1) The present invention proposes a transformer power consumption-temperature rise monitoring method based on regression confidence band. The method establishes a linear regression equation with power consumption change as independent variable and temperature rise as dependent variable. It also sets a dynamic confidence band with a preset multiple in combination with regression standard error, so that the alarm boundary can be automatically adjusted according to the actual load conditions. This fundamentally solves the problem that fixed threshold is prone to false alarms at low load and easy to miss alarms at high load. (2) By truncating historical data, filtering invalid data and verifying data volume, the robustness of the modeling sample and the statistical credibility of the regression results are effectively improved. Furthermore, a backup early warning mechanism for historical extreme values ​​is reserved when the regression model is unavailable, ensuring the continuity and compatibility of the monitoring system. (3) The regression parameters and confidence band parameters are stored in the database, which makes it easy for external monitoring systems to call them directly, thus realizing full-process automation from data preprocessing, dynamic modeling to anomaly judgment. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of a transformer energy-temperature rise monitoring method based on regression confidence bands. Detailed Implementation

[0014] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0015] This invention proposes a transformer energy consumption-temperature rise monitoring method based on regression confidence bands, applicable to transformer operation status monitoring and anomaly early warning scenarios in various power systems. This method analyzes the temperature rise characteristics of the transformer under different load conditions to establish dynamic alarm boundaries, thereby achieving accurate judgment of the transformer's operating status, such as... Figure 1 As shown, the method mainly includes the following steps: S1: Collect historical operating data of the transformer, which includes at least ambient temperature, equipment temperature and power readings; S2: Based on historical operating data, calculate the change in electricity consumption and the temperature rise value within a preset time window at each sampling moment, wherein the temperature rise value is the difference between the equipment temperature and the ambient temperature; S3: Preprocess historical operating data, including filtering invalid or abnormal data and truncating the temperature rise value corresponding to the same change in electricity to retain data with a preset intermediate ratio. S4: Using the change in electricity consumption in the preprocessed data as the independent variable and the temperature rise as the dependent variable, perform linear regression analysis to obtain the linear regression equation and the regression standard error; S5: Calculate the dynamic confidence band based on the regression standard error. The upper and lower limits of the dynamic confidence band are the predicted value of the linear regression equation plus and minus the regression standard error by a preset multiple, respectively. S6: Real-time acquisition of the transformer's current operating data, calculation of current electricity consumption change and current temperature rise; S7: Substitute the current change in electricity into the linear regression equation to obtain the predicted temperature rise value, and determine whether the current temperature rise value is within the dynamic confidence band. If yes, return to step S6; otherwise, determine that the transformer is operating abnormally and issue a warning message.

[0016] The present invention will now be described in detail with reference to specific implementation methods.

[0017] In practical applications, this monitoring method first requires acquiring historical operating data of the transformer as the basis for modeling. This historical operating data comes from various sensors and energy metering devices deployed at the transformer site, specifically including ambient temperature collected by ambient temperature sensors, equipment temperature collected by temperature sensors installed on the transformer body, and energy readings recorded by energy meters. During actual data acquisition, the sampling frequency of various data types can be set according to actual monitoring needs; for example, it can be set to collect data once per minute or once every five minutes. This invention does not strictly limit this. To ensure the accuracy of subsequent modeling, the collected historical operating data should cover the transformer's operating conditions under different load conditions. It is generally recommended to collect at least one week of continuous operating data to fully reflect the transformer's temperature rise characteristics under various load levels.

[0018] After obtaining the aforementioned historical operating data, the raw data needs to be preliminarily processed to transform it into characteristic parameters that reflect the transformer's operating status. Specifically, for each sampling moment, the change in electricity consumption within a preset time window prior to that moment, as well as the temperature rise at that moment, need to be calculated. The change in electricity consumption reflects the load change of the transformer within that time window, i.e., the degree of change in energy consumption; a larger value indicates a more drastic load change per unit time. In practical applications, the preset time window can be selected based on the transformer's thermal response characteristics, typically using a one-hour window. This is because a one-hour time scale can smooth out the interference of short-term fluctuations on modeling and better capture the pattern of transformer temperature rise with load changes. The change in electricity consumption is calculated by subtracting the electricity consumption reading one hour prior from the current electricity reading. To facilitate subsequent data processing and analysis, the change in electricity consumption is rounded to an integer. This allows data under similar load levels to be grouped into the same category, improving the stability of statistical analysis. Temperature rise is the difference between the equipment temperature and the ambient temperature. This value reflects the heat accumulation effect caused by internal losses during the operation of the transformer and is a key indicator for judging the operating status of the transformer.

[0019] After calculating the aforementioned characteristic parameters, data preprocessing is required to remove invalid and outlier samples from the original data, ensuring that the data used for modeling accurately reflects the temperature rise characteristics of the transformer under normal operating conditions. Preprocessing includes multiple levels of filtering. First, data format conversion and validity checks are performed. Since field-collected temperature data is typically transmitted and stored in hexadecimal format, the hexadecimal strings must be converted to decimal values. Data that fails to convert or exceeds a reasonable range is directly discarded. Second, preliminary filtering of temperature rise values ​​is conducted. Based on the physical characteristics of transformers, under normal circumstances, the temperature rise will not exceed a certain limit. For example, when the temperature rise reaches or exceeds 50°C, it usually indicates a sensor malfunction or that the transformer is under extreme abnormal operating conditions. Including such data in the modeling would severely distort the regression results; therefore, data with a temperature rise greater than or equal to 50°C must be excluded.

[0020] Furthermore, preprocessing includes truncating multiple temperature rise values ​​corresponding to the same change in electricity consumption. In actual operating data, for the same integer change in electricity consumption, the corresponding temperature rise values ​​often exhibit a certain degree of dispersion due to the influence of ambient temperature fluctuations, measurement errors, and other random factors. To eliminate the influence of extreme outliers on regression analysis, these temperature rise values ​​need to be screened. Specifically, for all temperature rise values ​​with the same change in electricity consumption, they are sorted according to their numerical value, and then a certain proportion of data at both ends of the sorting result is removed. For example, 10% of the data at each end is removed, and the middle 80% of the data is retained as the valid temperature rise sample for that change in electricity consumption. This truncation method can effectively filter out outliers caused by instantaneous sensor interference, data transmission errors, etc., making the data used in subsequent regression analysis more concentrated in reflecting the typical temperature rise characteristics under that load level.

[0021] After the above preprocessing, a valid dataset suitable for regression modeling is obtained. Each sample point in this dataset contains an integer change in electricity consumption and its corresponding temperature rise. Based on this, linear regression analysis is performed with the change in electricity consumption as the independent variable and the temperature rise as the dependent variable. Linear regression analysis can use the least squares method for parameter estimation, calculating statistics such as the slope, intercept, and standard error of the linear regression equation. The linear regression equation describes the average linear relationship between the change in electricity consumption and the temperature rise, i.e., the expected change in temperature rise when the change in electricity consumption increases by one unit. The standard error reflects the dispersion of the actual temperature rise relative to the predicted value; a larger standard error indicates a higher degree of dispersion of data points and relatively lower prediction accuracy of the regression model.

[0022] When performing linear regression analysis, it is essential to ensure that the amount of data used for modeling meets the basic requirements for statistical significance. Generally, the regression analysis results only have statistical significance when the number of preprocessed data samples is greater than or equal to a certain number (e.g., 10 samples). If the data volume is insufficient, the regression parameters cannot be reliably estimated. In this case, the system should skip the current regression modeling and continue using the previously successfully constructed linear regression equation and dynamic confidence band. This approach ensures that the monitoring system can maintain basic alarm functions in the initial stage when data accumulation is insufficient. As the running time increases and the data sample gradually becomes richer, the model will be continuously updated and optimized.

[0023] After obtaining the linear regression equation and regression standard error, a dynamic confidence band for anomaly detection can be constructed. The core idea of ​​the dynamic confidence band is that, under normal conditions, the temperature rise of the transformer should fall within a range close to the predicted regression value with a high probability. If the temperature rise deviates too far from the predicted value, it indicates that the transformer's operating condition may be abnormal. Specifically, the upper and lower limits of the dynamic confidence band are set as the predicted value of the linear regression equation plus a preset multiple of the regression standard error, and the predicted value minus the same multiple of the regression standard error, respectively. This preset multiple determines the width of the confidence band, thus affecting the alarm sensitivity. In a preferred embodiment of the invention, the preset multiple is set to 2.0. At this value, the dynamic confidence band corresponds to approximately 95% confidence level, meaning that the temperature rise under normal operating conditions has a 95% probability of falling within this confidence band range. If higher alarm sensitivity is required in practical applications, the preset multiple can be appropriately reduced; if a lower false alarm rate is required, the preset multiple can be appropriately increased. This parameter can be flexibly adjusted based on field operating experience and actual needs.

[0024] After completing the modeling and confidence band calculation, the system enters the real-time monitoring phase. During real-time monitoring, the system continuously collects the transformer's current operating data, including ambient temperature, equipment temperature, and energy readings, and calculates the current energy change and temperature rise using the same method as historical data processing. The current energy change is then substituted into the established linear regression equation to calculate the predicted temperature rise at the current load level. The system then determines whether the actual temperature rise falls within the dynamic confidence band. If the current temperature rise is between the upper and lower limits of the confidence band, it indicates that the transformer's current temperature rise conforms to historical statistical patterns and is under normal operating conditions. The system continues to collect data for the next moment for monitoring. If the current temperature rise exceeds the upper and lower limits of the confidence band, it indicates an abnormal deviation of the transformer's current temperature rise relative to its load level, possibly due to internal transformer faults, deteriorated heat dissipation, or other abnormal reasons. In this case, the system immediately issues a warning to notify maintenance personnel for inspection and handling.

[0025] In practical engineering applications, to facilitate integration with other monitoring systems, this invention also provides a model parameter storage function. The system stores the regression parameters of each successfully constructed linear regression equation, including slope and intercept, as well as statistical measures such as regression standard error, preset multiples used for dynamic confidence bands, and goodness of fit, into the database. This allows external monitoring systems to access the database to obtain the currently valid alarm model parameters, thereby achieving unified push and display of early warning information. Since the database stores the latest updated model parameters, external systems do not need to repeatedly perform modeling calculations; they can directly access the database to obtain the dynamic alarm boundaries.

[0026] Furthermore, considering potential special circumstances during actual operation, such as insufficient historical data accumulation preventing effective linear regression analysis, or the loss of model parameters in the database, this invention also designs a backup early warning method to ensure the reliability of the monitoring system. This backup early warning method constructs alarm boundaries based on historical extreme temperature rise values. Specifically, the system pre-calculates and stores the historical maximum and minimum temperature rise values ​​corresponding to each integer change in electricity consumption. When the primary early warning method fails, the system switches to the backup method, using the historical extreme temperature rise value corresponding to the current change in electricity consumption as the alarm boundary. That is, if the current temperature rise value exceeds the maximum or minimum value range of the change in electricity consumption in historical records, it is determined to be abnormal. Although this backup scheme is less accurate than the dynamic confidence band method, it can maintain basic monitoring functions under special circumstances, improving the system's fault tolerance and applicability.

[0027] In practical implementation, the monitoring method of this invention can be deployed based on existing data acquisition and monitoring platforms. The data acquisition component can reuse existing sensors and data acquisition devices in substations or distribution networks, requiring no additional hardware investment. The data processing and modeling component can use servers or industrial control computers as computing nodes to run corresponding data processing and regression analysis programs. The real-time monitoring and early warning component can interface with existing monitoring systems to push early warning information to the monitoring terminals of maintenance personnel. The deployment and implementation of the entire system can be flexibly adjusted according to actual site conditions, exhibiting good scalability and adaptability.

[0028] Through the above technical solution, the transformer power consumption-temperature rise monitoring method proposed in this invention, based on regression confidence bands, uses a linear regression model to describe the statistical relationship between power consumption changes and temperature rise values. A dynamic confidence band is constructed based on the regression standard error as the boundary for anomaly detection, enabling automatic adjustment and adaptation of the alarm threshold. Compared with traditional monitoring methods using fixed temperature rise thresholds, this method can dynamically adjust the alarm boundary according to the actual load level of the transformer, avoiding overly sensitive false alarms at low loads and ensuring effective capture of abnormal temperature rises at high loads, significantly improving the accuracy and reliability of monitoring. Simultaneously, this method performs multi-level preprocessing and truncation of historical data, effectively eliminating interference from abnormal data on modeling, making the regression model more robust. Furthermore, the establishment of a backup early warning mechanism further enhances the system's fault tolerance and applicability, ensuring effective monitoring protection for the safe operation of the transformer under various conditions.

[0029] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0030] Furthermore, in this invention, descriptions involving terms such as "first," "second," and "a" are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0031] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0032] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

Claims

1. A method for monitoring transformer energy consumption and temperature rise based on regression confidence bands, characterized in that, Including the following steps: S1: Collect historical operating data of the transformer, which includes at least ambient temperature, equipment temperature and power readings; S2: Based on historical operating data, calculate the change in electricity consumption and the temperature rise value within a preset time window at each sampling moment, wherein the temperature rise value is the difference between the equipment temperature and the ambient temperature; S3: Preprocess historical operating data, including filtering invalid or abnormal data and truncating the temperature rise value corresponding to the same change in electricity to retain data with a preset intermediate ratio. S4: Using the change in electricity consumption in the preprocessed data as the independent variable and the temperature rise as the dependent variable, perform linear regression analysis to obtain the linear regression equation and the regression standard error; S5: Calculate the dynamic confidence band based on the regression standard error, wherein the upper and lower limits of the dynamic confidence band are the predicted value of the linear regression equation plus and minus the regression standard error by a preset multiple, respectively. S6: Real-time acquisition of the transformer's current operating data, calculation of current electricity consumption change and current temperature rise; S7: Substitute the current change in electricity into the linear regression equation to obtain the predicted temperature rise value, and determine whether the current temperature rise value is within the dynamic confidence band. If yes, return to step S6; otherwise, determine that the transformer is operating abnormally and issue a warning message.

2. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, In step S2, the preset time window is 1 hour, the change in electricity consumption is the change in electricity consumption reading within 1 hour, and the change in electricity consumption is rounded to an integer.

3. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, In step S3, the temperature rise value corresponding to the same change in electricity is truncated to retain data at a preset intermediate ratio. Specifically: For multiple temperature rise values ​​with the same change in electrical energy, sort them by numerical value, remove the data at both ends with a preset ratio, and retain the data in the middle as the effective temperature rise data of the change in electrical energy.

4. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, In step S3, the preprocessing also includes: Historical operating data with a temperature rise greater than or equal to 50°C are filtered out to exclude invalid samples from sensor malfunctions or extreme abnormal operating conditions.

5. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, In step S4, linear regression analysis is performed when the amount of preprocessed data is greater than or equal to a preset amount; otherwise, the regression analysis is skipped and the previously successfully constructed linear regression equation and dynamic confidence band are used.

6. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, In step S5, the preset multiple is 2.0, and the corresponding dynamic confidence band is the linear regression predicted value ± 2 times the regression standard error. The dynamic confidence band corresponds to a 95% confidence level.

7. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, It also includes the following steps: S8: Store the regression parameters, regression standard error, preset multiples of dynamic confidence band, and goodness of fit of the linear regression equation into the database for use by the external monitoring system for abnormal early warning.

8. The transformer power consumption-temperature rise monitoring method based on regression confidence band as described in claim 1, characterized in that, It also includes backup early warning methods: In step S1, the historical maximum and minimum temperature rise values ​​corresponding to each integer change in electricity are pre-calculated and stored; In step S4, when linear regression analysis cannot be performed or the dynamic confidence band is unavailable, the historical extreme value of temperature rise corresponding to the current change in electricity consumption is used as the alarm boundary.

Images