Secondary circuit defect intelligent detection method based on infrared temperature measurement
By determining the periodic characteristics of the initial detrending component in STL decomposition, adaptive seasonal correction and double decomposition are performed, solving the problem of abnormal signals being masked in secondary loop detection and achieving higher detection accuracy and sensitivity.
Patent Information
- Application Number
- CN202511840968.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-09
- Publication Date
- 2026-02-06
AI Technical Summary
In existing technologies for secondary circuit defect detection, the temperature data variation patterns caused by seasonal corrections are not significant and are easily changed, which leads to the masking or weakening of abnormal signals, affecting the detection accuracy and sensitivity.
By obtaining the seasonal uncertainty factor of the initial detrended component, it is determined whether it has a clear periodic characteristic. Seasonal correction and double decomposition operations are only performed when there is no clear periodic characteristic. The optimal residual component is adaptively obtained, reducing the computational cost and preserving the true abnormal signal.
It improves the accuracy and sensitivity of secondary circuit defect detection, preserves the true abnormal signals to the greatest extent, and reduces computational costs.
Smart Images
Figure CN121476800A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to an intelligent detection method for secondary circuit defects based on infrared thermometry. Background Technology
[0002] Secondary circuits are low-voltage, low-current circuits in a power system used for protection, measurement, control, and signal transmission. They do not carry main current but achieve protection and control functions by monitoring signals from the primary circuit. To prevent equipment damage, personnel injuries, and system instability caused by circuit faults, defect detection of secondary circuits is necessary to ensure the safe and reliable operation of the power system.
[0003] Existing technologies typically utilize infrared thermometry to acquire temperature time-series data in the secondary circuit. Then, the temperature time-series data is decomposed using STL to obtain seasonal components, trend components, and residual components. The residual components can reflect the characteristics of short-term abnormal fluctuations or sudden events, and are used as the main basis for defect detection in the secondary circuit.
[0004] In the STL decomposition process, seasonal corrections are performed during iterations to improve the accuracy and reliability of time-series data analysis, enabling more precise separation of trend and seasonal components. This facilitates targeted feature extraction and anomaly identification for different components. However, the instability of contact resistance and the non-fixed nature of current carrying capacity in the secondary circuit can make temperature data variations insignificant and volatile. Consequently, in each iteration, the corrected seasonal components may misfit the seasonal components or become overly smoothed, weakening the true fluctuations of trend and residual components. Ultimately, this may mask or weaken anomalous signals after decomposition, leading to inaccurate anomaly detection results.
[0005] Therefore, how to evaluate the impact of seasonal correction on the decomposition results of temperature time series data, adaptively obtain the optimal decomposition results, and improve the accuracy and sensitivity of defect detection in secondary circuits has become an urgent problem to be solved. Summary of the Invention
[0006] In view of this, embodiments of the present invention provide an intelligent detection method for secondary circuit defects based on infrared thermometry, in order to solve the problem of how to evaluate the impact of seasonal correction on the decomposition results of temperature time series data, adaptively obtain the optimal decomposition results, and improve the accuracy and sensitivity of secondary circuit defect detection.
[0007] This invention provides an intelligent detection method for secondary circuit defects based on infrared thermometry, which includes the following steps:
[0008] Acquire the temperature data of any monitoring point in the secondary circuit of the power system at each moment, and combine the temperature data within a preset time period including the current moment into a temperature data sequence;
[0009] In any inner loop iteration of STL decomposition of the temperature data sequence, an initial detrended component is obtained, a fast Fourier transform is performed on the initial detrended component to obtain a frequency domain signal, the seasonal uncertainty factor of the initial detrended component is obtained according to the energy distribution characteristics of the frequency domain signal, and the seasonal uncertainty factor is used to determine whether the initial detrended component has a clear periodic characteristic.
[0010] If the initial detrending component does not have a clear periodic characteristic, then the initial detrending component is subjected to seasonal correction to obtain a corrected seasonal component, and the initial detrending component is recorded as the uncorrected seasonal component. Based on the corrected seasonal component and the uncorrected seasonal component, the target residual component of any inner loop iteration is obtained.
[0011] Repeat the above-described method for obtaining the target residual component in any inner loop iteration. After the convergence condition of any inner loop is met, enter the outer loop of STL decomposition until the convergence condition of the outer loop is met, and obtain the final residual component of the temperature data sequence. Perform defect detection on any monitoring point in the secondary circuit of the power system at the current moment.
[0012] Preferably, obtaining the seasonal uncertainty factor of the initial detrended component based on the energy distribution characteristics of the frequency domain signal includes:
[0013] The dominant frequency component in the frequency domain signal and its frequency components within a preset neighborhood range are recorded as the main frequency components. The proportion of the total energy of all main frequency components in the total energy of the frequency domain signal is obtained to obtain the main energy proportion. The absolute value of the difference between the main energy proportion and the constant 0.5 is obtained. The product of the absolute value of the difference and the constant 2 is calculated. The difference between the constant 1 and the product is obtained to obtain the seasonal uncertainty factor of the initial detrending component.
[0014] Preferably, determining whether the initial detrending component has a clear cyclical characteristic based on the seasonal uncertainty factor includes:
[0015] Set a seasonal uncertainty factor threshold. If the seasonal uncertainty factor of the initial detrending component is greater than the seasonal uncertainty factor threshold, then it is confirmed that the initial detrending component does not have a clear periodic characteristic.
[0016] If the seasonal uncertainty factor of the initial detrending component is less than or equal to the seasonal uncertainty factor threshold, then the initial detrending component is confirmed to have a clear periodic characteristic.
[0017] Preferably, obtaining the target residual component for any inner loop iteration based on the corrected seasonal component and the uncorrected seasonal component includes:
[0018] Obtain the residual components corresponding to the corrected seasonal component and the uncorrected seasonal component respectively. For any residual component, obtain the decomposition effect score of the residual component based on the data distribution characteristics of the residual component.
[0019] Obtain the decomposition effect score for each residual component, and take the residual component corresponding to the maximum decomposition effect score as the target residual component for any inner loop iteration.
[0020] Preferably, obtaining the decomposition effect score of any residual component based on the data distribution characteristics of any residual component includes:
[0021] Based on the difference in data distribution characteristics between any residual component and the standard normal distribution, the degree of deviation of any residual component from the normal distribution is obtained;
[0022] Obtain outliers in any residual component, and based on the data distribution characteristics of outliers in any residual component, obtain the degree of concentration of outliers in any residual component;
[0023] The sum of the deviation from the normal distribution and the concentration of outliers is obtained to get the decomposition anomaly degree of any residual component. The ratio of the decomposition anomaly degree to constant 4 is calculated, and the difference between constant 1 and the ratio is obtained to get the decomposition effect score of any residual component.
[0024] Preferably, obtaining the degree of deviation from the normal distribution of any residual component based on the difference in data distribution characteristics between any residual component and the standard normal distribution includes:
[0025] Obtain the cumulative distribution function of any residual component and the standard normal distribution respectively, calculate the absolute value of the difference between the cumulative distribution function of any residual component and the cumulative distribution function of the standard normal distribution, obtain the absolute value function of the difference, and normalize the integral result of the absolute value function of the difference within its preset range to obtain the first normal distribution deviation degree;
[0026] Obtain the absolute value of the difference between the maximum value of any residual component and the constant 1 to obtain the degree of deviation of the maximum value; calculate the ratio of the degree of deviation of the maximum value to the minimum value of any residual component to obtain the relative deviation value; obtain the difference between the constant 1 and the relative deviation value to obtain the degree of deviation of the second normal distribution.
[0027] The degree of deviation from the normal distribution of any residual component is obtained by adding the first degree of deviation from the normal distribution to the second degree of deviation from the normal distribution.
[0028] Preferably, the step of obtaining the degree of outlier concentration of any residual component based on the data distribution characteristics of outliers in any residual component includes:
[0029] Obtain the percentage of all outliers in any residual component to obtain the first outlier concentration.
[0030] The outliers in any residual component are formed into an outlier sequence. The time interval between every two adjacent outliers in the outlier sequence is obtained, and the average time interval is obtained. The negative of the average time interval is substituted into an exponential function with the natural constant as the base to obtain the second outlier concentration.
[0031] The degree of outlier concentration of any residual component is obtained by adding the first degree of outlier concentration to the second degree of outlier concentration.
[0032] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:
[0033] In this invention, the seasonal uncertainty factor of the initial detrended component is obtained to determine whether the initial detrended component has a clear periodic characteristic. Subsequent calculations are only performed when the initial detrended component does not have a clear periodic characteristic, thereby minimizing computational costs. Based on the corrected and uncorrected seasonal components, the target residual component for any inner loop iteration is adaptively obtained. By selectively choosing the residual component, the true abnormal signal can be preserved to the greatest extent, enabling effective detection of temperature data at any monitoring point in the secondary loop and improving the accuracy and sensitivity of defect detection in the secondary loop. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] Figure 1 This is a flowchart of a method for intelligent detection of secondary circuit defects based on infrared thermometry, provided in Embodiment 1 of the present invention. Detailed Implementation
[0036] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.
[0037] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.
[0038] To illustrate the technical solution of the present invention, specific embodiments are described below.
[0039] See Figure 1 This is a flowchart of a method for intelligent detection of secondary circuit defects based on infrared thermometry, as provided in Embodiment 1 of the present invention. Figure 1 As shown, the method may include:
[0040] Step S101: Obtain the temperature data of any monitoring point in the secondary circuit of the power system at each time, and form a temperature data sequence by combining the temperature data within a preset time period including the current time.
[0041] To prevent equipment damage, casualties, and system instability caused by secondary circuit faults in power systems, existing technologies typically utilize infrared thermography to acquire temperature time-series data in the secondary circuits. Then, the temperature time-series data is decomposed using STL to obtain seasonal components, trend components, and residual components. The residual components can reflect the characteristics of short-term abnormal fluctuations or sudden events. Using the residual components as the main basis, defect detection in the secondary circuits is carried out to ensure the safe and reliable operation of the power system.
[0042] In this embodiment, infrared temperature measurement equipment (such as a fixed infrared sensor or thermal imager array) is used to continuously monitor the temperature of key components such as secondary circuit terminals, relay contacts, and transformer leads, obtaining temperature data for each monitoring point (i.e., key component) at each moment. Since the method for defect detection at each monitoring point in the secondary circuit is the same, this embodiment takes any monitoring point as an example, acquiring the temperature data of any monitoring point in the secondary circuit of the power system at each moment. The temperature data within a preset time period, including the current moment, is combined into a temperature data sequence for defect detection at any monitoring point. In this embodiment, the temperature data acquisition frequency is set to 5Hz, and the preset time period is set to 2 hours. This is not limited and can be set according to the specific implementation scenario.
[0043] The acquired temperature data needs to be preprocessed to form a temperature data sequence. Data preprocessing is an existing technology, which will be briefly described here: (1) Using the box plot method or (1) Mark extreme outliers in the temperature data as noise data points; (2) Use linear interpolation or local weighted regression to complete the noise data points; (3) Use time window averaging or spline difference to complete the missing points; (4) Use wavelet denoising to reduce the interference of ambient temperature and measurement error.
[0044] In the inner loop of the STL decomposition process, seasonal corrections are performed during iterations to improve the accuracy and reliability of time-series data analysis. This allows for more precise separation of trend and seasonal components, facilitating targeted feature extraction and anomaly identification for different components. However, temperature in the secondary loop is primarily affected by contact resistance and current carrying capacity. Temperature changes are mainly caused by variations in contact resistance and current-carrying heating effects of components such as wires, terminals, and relays. The instability of contact resistance and the non-fixed nature of current carrying capacity in the secondary loop make the temperature data variation patterns insignificant and volatile. Consequently, in each iteration, the corrected seasonal components may misfit the seasonal components or become overly smoothed, weakening the true fluctuations of trend and residual components. Ultimately, this may mask or weaken the anomalous signals after decomposition, leading to inaccurate anomaly detection results.
[0045] Therefore, in this embodiment of the invention, the temperature data sequence of the secondary loop is obtained. In any inner loop iteration of the STL decomposition of the temperature data sequence, the seasonal uncertainty factor of the initial detrended component is obtained, and it is determined whether the initial detrended component has a clear periodic characteristic. When the initial detrended component does not have a clear periodic characteristic, the target residual component of any inner loop iteration is adaptively obtained, and then the STL decomposition of the temperature data sequence is completed to obtain the final residual component. The secondary loop is then used for defect detection, thereby improving the accuracy and sensitivity of defect detection.
[0046] Step S102: In any inner loop iteration of STL decomposition of the temperature data sequence, an initial detrended component is obtained, a fast Fourier transform is performed on the initial detrended component to obtain a frequency domain signal, the seasonal uncertainty factor of the initial detrended component is obtained according to the energy distribution characteristics of the frequency domain signal, and the initial detrended component is determined to have a clear periodic characteristic based on the seasonal uncertainty factor.
[0047] STL decomposition consists of an inner loop and an outer loop. The inner loop is responsible for progressively refining trend and seasonal components, while the outer loop reduces the impact of outliers through robust weights. The outer loop typically refers to the outermost loop, which controls the number of iterations of the entire operation. For example, if multiple operations need to be performed on each element in a container, then this loop controlling the entire process is the outer loop. The inner loop is a loop nested within the outer loop, performing multiple operations on each element in the current iteration of the outer loop.
[0048] In the inner loop iteration of STL decomposition of temperature data sequence, in order to reduce the possible problems of the corrected seasonal components in the inner loop, a double decomposition operation is required. That is, on the basis of the seasonal correction in the traditional STL decomposition, a second seasonal non-correction operation is performed to obtain two residual components. The target residual component is selected from the two residual components to provide a better data source for subsequent defect detection. However, this double decomposition operation will increase the amount of computation and computational cost.
[0049] Because temperature data in the secondary loop may exhibit insignificant and volatile patterns, the periodicity of the initial detrending component in each iteration of the inner loop may be unclear. Furthermore, the period length, amplitude, or phase may change over time, reducing the applicability of the seasonal component extracted in the previous iteration to the current iteration. This can lead to fluctuations in the seasonal component updated in each iteration, making it difficult to accurately separate the trend component from the residual component. Therefore, when the initial detrending component has a clear periodic characteristic, the corrected seasonal component can be extracted with a clear period, eliminating the need for a double decomposition operation; traditional STL decomposition can be performed.
[0050] Therefore, to minimize computational load, this embodiment obtains the initial detrended component in any inner loop iteration of the STL decomposition of the temperature data sequence. By analyzing the initial detrended component, the double decomposition operation is performed only when the initial detrended component does not have a clear periodic characteristic. This minimizes computational cost while obtaining the optimal decomposition result (target residual component). For example, in the k-th iteration, the initial detrended component of the k-th iteration... Where Y is the temperature data sequence, This is the trend component obtained in the (k-1)th iteration.
[0051] To better analyze the periodic characteristics of the initial detrended component, this embodiment performs a Fast Fourier Transform (FFT) on the initial detrended component to obtain a frequency domain signal. The FFT is an existing technology and will not be elaborated here. Based on the energy distribution characteristics of the frequency domain signal, the seasonal uncertainty factor of the initial detrended component is obtained to determine whether the initial detrended component has a clear periodic characteristic.
[0052] The method for obtaining the seasonal uncertainty factor of the initial detrended component based on the energy distribution characteristics of the frequency domain signal is as follows:
[0053] The dominant frequency component in the frequency domain signal and its frequency components within a preset neighborhood range are denoted as the main frequency components. In this embodiment, the preset neighborhood range is set to 10% of the relative bandwidth of the dominant frequency, which is the conventional neighborhood setting range. There is no limitation here, and it can be set according to the specific implementation scenario. The proportion of the total energy of all main frequency components in the total energy of the frequency domain signal is obtained to obtain the main energy proportion. The absolute value of the difference between the main energy proportion and the constant 0.5 is obtained. The product of the absolute value of the difference and the constant 2 is calculated. The difference between the constant 1 and the product is obtained to obtain the seasonal uncertainty factor of the initial detrended component.
[0054] In one implementation, the formula for calculating the seasonal uncertainty factor of the initial detrending component is:
[0055]
[0056] Where U is the seasonal uncertainty factor of the initial detrending component; This represents the total energy of all major frequency components in the frequency domain signal. This represents the total energy of the frequency domain signal.
[0057] It should be noted that, The dominant energy percentage reflects the proportion of energy near the dominant periodic frequency in a frequency domain signal to the total energy. The larger the value, the stronger the periodicity of the initial detrending component. The smaller the value, the weaker the periodicity of the initial detrending component. Strong or weak periodicity both indicate that the initial detrending component has a clear periodic characteristic. The closer U is to 0.5, the more ambiguous the periodic characteristics of the initial detrending component become, and the more uncertain its periodicity. In other words, the larger U is, the more ambiguous the periodic characteristics of the initial detrending component become, and the more uncertain its periodicity becomes.
[0058] Furthermore, based on the seasonal uncertainty factor, it is determined whether the initial detrended component has a clear periodic characteristic. Specifically, a seasonal uncertainty factor threshold is set. Setting the threshold too high may prevent some cycles with moderate uncertainty from being identified as requiring double decomposition, resulting in the residuals failing to accurately reflect anomalies; setting it too low may trigger double decomposition even with slightly higher seasonal uncertainty, significantly increasing computational costs. Therefore, in this embodiment, the seasonal uncertainty factor threshold is set to 0.7, but this is not a limitation and can be set according to the specific implementation scenario. If the seasonal uncertainty factor of the initial detrended component is greater than 0.7, it is confirmed that the initial detrended component does not have a clear periodic characteristic. If traditional STL decomposition is used for seasonal term correction, the corrected seasonal component may have problems, requiring a double decomposition operation. If the seasonal uncertainty factor of the initial detrended component is less than or equal to 0.7, it is confirmed that the initial detrended component has a clear periodic characteristic. Traditional STL decomposition is used for seasonal term correction, and the corresponding residual component is used as the target residual component for any current inner loop iteration.
[0059] Thus, we obtain the result of whether the initial detrending component has a clear periodic characteristic.
[0060] Step S103: If the initial detrending component does not have a clear periodic characteristic, then the initial detrending component is subjected to seasonal correction processing to obtain a corrected seasonal component, and the initial detrending component is recorded as the uncorrected seasonal component. Based on the corrected seasonal component and the uncorrected seasonal component, the target residual component of any inner loop iteration is obtained.
[0061] If the initial detrending component does not have a clear periodic characteristic, a double decomposition operation needs to be performed on the initial detrending component. Seasonal correction processing is performed on the initial detrending component to obtain the corrected seasonal component. The initial detrending component is recorded as the uncorrected seasonal component. The residual components corresponding to the corrected seasonal component and the uncorrected seasonal component are obtained respectively. By analyzing the decomposition effect of each residual component, the residual component with the best decomposition effect is selected as the target residual component of the current inner loop iteration.
[0062] When the initial detrending component fails to accurately reflect the periodicity, the effect of seasonal correction may be either good or bad. Excessive or insufficient seasonal correction can alter the allocation of seasonal components in the original data. For example, seasonal patterns may remain in the residuals, masking true anomalous signals, or important seasonal information may be removed, leading to increased noise. These factors affect the representation of anomalous signals in the residuals relative to the original data, causing the residual components to deviate from a normal distribution and exhibiting a concentration of outliers. Therefore, for any given residual component, a decomposition performance score can be obtained based on its data distribution characteristics to reflect the decomposition effect.
[0063] The method for obtaining the decomposition effect score of any residual component based on the data distribution characteristics of any residual component is as follows:
[0064] (1) Based on the difference in data distribution characteristics between any residual component and the standard normal distribution, obtain the degree of deviation of any residual component from the normal distribution.
[0065] Specifically, the cumulative distribution function of any residual component and the standard normal distribution are obtained respectively. The standard normal distribution and the cumulative distribution function are existing technologies and will not be described in detail here. The absolute value of the difference between the cumulative distribution function of any residual component and the cumulative distribution function of the standard normal distribution is calculated to obtain the absolute value function of the difference. The integral result of the absolute value function of the difference within its preset range is normalized to obtain the first normal distribution deviation degree.
[0066] Obtain the absolute value of the difference between the maximum value of any residual component and the constant 1 to obtain the degree of deviation of the maximum value; calculate the ratio of the degree of deviation of the maximum value to the minimum value of any residual component to obtain the relative deviation value; obtain the difference between the constant 1 and the relative deviation value to obtain the degree of deviation of the second normal distribution.
[0067] The degree of deviation from the normal distribution of any residual component is obtained by adding the first degree of deviation from the normal distribution to the second degree of deviation from the normal distribution.
[0068] In one embodiment, taking the i-th residual component as an example, the formula for calculating the degree of deviation from the normal distribution of the i-th residual component is:
[0069]
[0070] in, The degree of deviation of the i-th residual component from the normal distribution; Let be the cumulative distribution function of the i-th residual component; The cumulative distribution function of the standard normal distribution; The maximum value of the i-th residual component; It is the minimum value of the i-th residual component; This is the normalization function; It is the absolute value symbol; It is the symbol for negative infinity; The symbol is positive infinity. In this embodiment, the preset range is... This means integrating the absolute value of the difference function over its domain. There are no restrictions here, and it can be set according to the specific implementation scenario.
[0071] It should be noted that, This represents the degree of deviation from the first normal distribution. The larger the value, the greater the difference between the data distribution structure of the i-th residual component and the data distribution structure of the standard normal distribution. The larger it is; This represents the degree of deviation from the second normal distribution, since the standard normal distribution function ranges from [0,1]. This indicates the degree of deviation of the i-th residual component from the maximum value in the standard normal distribution. (can be regarded as) The expression represents the degree of deviation of the i-th residual component from the minimum value in the standard normal distribution. This reflects the relative deviation of the i-th residual component from the standard normal distribution at extreme values. The closer the value is to 1, the more similar the deviation of the i-th residual component is to the extreme values in the standard normal distribution, and the closer the data characteristics of the i-th residual component are to the standard normal distribution. The smaller it is.
[0072] Since the integral result of the absolute difference function within its preset range is essentially an average quantification of the differences in the global distribution, it may lack sufficient sensitivity to extreme values. Therefore, when using... It reflects the relative deviation of the i-th residual component from the standard normal distribution at extreme values, highlights the amplitude characteristics of extreme anomalies in the residual component, and thus more comprehensively assesses the degree of deviation of the i-th residual component from the normal distribution.
[0073] (2) Obtain outliers in any residual component, and obtain the degree of concentration of outliers in any residual component based on the data distribution characteristics of outliers in any residual component.
[0074] Specifically, the proportion of all outliers in any residual component is obtained to determine the first outlier concentration. The determination of outliers in the residual component is determined according to actual needs and scenarios (for example, box plot method can be used), and there are no restrictions here.
[0075] The outliers in any residual component are formed into an outlier sequence. The time interval between every two adjacent outliers in the outlier sequence is obtained, and the average time interval is obtained. The negative of the average time interval is substituted into an exponential function with the natural constant as the base to obtain the second outlier concentration.
[0076] The degree of outlier concentration of any residual component is obtained by adding the first degree of outlier concentration to the second degree of outlier concentration.
[0077] In one implementation, taking the i-th residual component as an example, the formula for calculating the degree of outlier concentration of the i-th residual component is:
[0078]
[0079] in, denoted as the concentration of outliers in the i-th residual component; m is the number of outliers in the i-th residual component; M is the total amount of data in the i-th residual component. The time corresponding to the (j+1)th outlier in the i-th outlier sequence; The time corresponding to the j-th outlier in the i-th outlier sequence; It is an exponential function with the natural constant as the base.
[0080] It should be noted that, This represents the concentration of outliers, which is the percentage of all outliers in the i-th residual component. The larger the value, the more outliers there are in the i-th residual component; The degree of concentration of the second outlier. This represents the average time interval between two adjacent outliers in the i-th residual component. The more outliers in the i-th residual component, and the shorter the average time interval between adjacent outliers, the more concentrated the outliers are in the i-th residual component. The larger it is.
[0081] (3) Based on the degree of deviation from the normal distribution and the degree of concentration of outliers of any residual component, obtain the decomposition effect score of any residual component.
[0082] Specifically, the sum of the deviation from the normal distribution and the concentration of outliers is obtained to get the decomposition anomaly degree of any residual component. The ratio of the decomposition anomaly degree to constant 4 is calculated, and the difference between constant 1 and the ratio is obtained to get the decomposition effect score of any residual component.
[0083] In one implementation, taking the i-th residual component as an example, the formula for calculating the decomposition effect score of the i-th residual component is:
[0084]
[0085] in, The decomposition effect score for the i-th residual component; The degree of deviation of the i-th residual component from the normal distribution; The degree of outlier concentration for the i-th residual component.
[0086] It should be noted that, and The smaller the value, the more the data of the i-th residual component conforms to a normal distribution, and the fewer outliers there are, with no clustered occurrences, indicating a better decomposition result. The larger it is.
[0087] According to the above method for obtaining the decomposition effect score of the i-th residual component, the decomposition effect score of each residual component is obtained, and the residual component corresponding to the maximum decomposition effect score is taken as the target residual component of any outer loop iteration.
[0088] Thus, the target residual component of any one inner loop iteration is obtained.
[0089] Step S104: Repeat the above-described method for obtaining the target residual component of any inner loop iteration. After satisfying the convergence condition of any inner loop, enter the outer loop of STL decomposition until the convergence condition of the outer loop is satisfied, and obtain the final residual component of the temperature data sequence. Perform defect detection on any monitoring point in the secondary circuit of the power system at the current moment.
[0090] Furthermore, the method for obtaining the target residual component in any one of the inner loop iterations described above is repeated until the convergence condition of any one of the inner loops is met, and the inner loop is completed. Then, the outer loop for STL decomposition of the temperature data sequence is entered, and robustness adjustment is performed. The outer loop for STL decomposition is repeated until the final decomposition condition of the outer loop is met, and the final residual component of the temperature data sequence is obtained. Defect detection is performed on any monitoring point in the secondary circuit of the power system at the current moment. Similarly, the final residual component of the temperature data sequence of each monitoring point in the secondary circuit is obtained, and defect detection is performed on each monitoring point in the secondary circuit of the power system. Finally, the detection result of the secondary circuit of the power system at the current moment is obtained.
[0091] It is worth noting that the focus of this embodiment of the invention lies in the inner loop iteration of the STL decomposition process. If the initial detrending component in the inner loop iteration does not have a clear periodic characteristic, a double decomposition operation is performed on the initial detrending component to select the target residual component in the inner loop iteration. The inner and outer loops of STL decomposition, as well as the defect detection using the final residual component, are existing technologies and will not be elaborated here.
[0092] In summary, in this embodiment of the invention, the seasonal uncertainty factor of the initial detrended component is obtained to determine whether the initial detrended component has a clear periodic characteristic. Subsequent calculations are only performed when the initial detrended component does not have a clear periodic characteristic, thereby minimizing computational costs. Based on the corrected and uncorrected seasonal components, the target residual component for any inner loop iteration is adaptively obtained. By selectively choosing the residual component, the true abnormal signal can be preserved to the greatest extent, enabling effective detection of temperature data at any monitoring point in the secondary loop and improving the accuracy and sensitivity of defect detection in the secondary loop.
[0093] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for intelligent detection of secondary circuit defects based on infrared thermometry, characterized in that, The aforementioned intelligent detection method for secondary circuit defects based on infrared thermometry includes: Acquire the temperature data of any monitoring point in the secondary circuit of the power system at each moment, and combine the temperature data within a preset time period including the current moment into a temperature data sequence; In any inner loop iteration of STL decomposition of the temperature data sequence, an initial detrended component is obtained, a fast Fourier transform is performed on the initial detrended component to obtain a frequency domain signal, the seasonal uncertainty factor of the initial detrended component is obtained according to the energy distribution characteristics of the frequency domain signal, and the seasonal uncertainty factor is used to determine whether the initial detrended component has a clear periodic characteristic. If the initial detrending component does not have a clear periodic characteristic, then the initial detrending component is subjected to seasonal correction to obtain a corrected seasonal component, and the initial detrending component is recorded as the uncorrected seasonal component. Based on the corrected seasonal component and the uncorrected seasonal component, the target residual component of any inner loop iteration is obtained. Repeat the above-described method for obtaining the target residual component in any inner loop iteration. After the convergence condition of any inner loop is met, enter the outer loop of STL decomposition until the convergence condition of the outer loop is met, and obtain the final residual component of the temperature data sequence. Perform defect detection on any monitoring point in the secondary circuit of the power system at the current moment.
2. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 1, characterized in that, The step of obtaining the seasonal uncertainty factor of the initial detrended component based on the energy distribution characteristics of the frequency domain signal includes: The dominant frequency component in the frequency domain signal and its frequency components within a preset neighborhood range are recorded as the main frequency components. The proportion of the total energy of all main frequency components in the total energy of the frequency domain signal is obtained to obtain the main energy proportion. The absolute value of the difference between the main energy proportion and the constant 0.5 is obtained. The product of the absolute value of the difference and the constant 2 is calculated. The difference between the constant 1 and the product is obtained to obtain the seasonal uncertainty factor of the initial detrending component.
3. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 1, characterized in that, The step of determining whether the initial detrending component has a clear cyclical characteristic based on the seasonal uncertainty factor includes: Set a seasonal uncertainty factor threshold. If the seasonal uncertainty factor of the initial detrending component is greater than the seasonal uncertainty factor threshold, then it is confirmed that the initial detrending component does not have a clear periodic characteristic. If the seasonal uncertainty factor of the initial detrending component is less than or equal to the seasonal uncertainty factor threshold, then the initial detrending component is confirmed to have a clear periodic characteristic.
4. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 1, characterized in that, The step of obtaining the target residual component for any inner loop iteration based on the corrected seasonal component and the uncorrected seasonal component includes: Obtain the residual components corresponding to the corrected seasonal component and the uncorrected seasonal component respectively. For any residual component, obtain the decomposition effect score of the residual component based on the data distribution characteristics of the residual component. Obtain the decomposition effect score for each residual component, and take the residual component corresponding to the maximum decomposition effect score as the target residual component for any inner loop iteration.
5. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 4, characterized in that, The step of obtaining the decomposition effect score of any residual component based on the data distribution characteristics of any residual component includes: Based on the difference in data distribution characteristics between any residual component and the standard normal distribution, the degree of deviation of any residual component from the normal distribution is obtained; Obtain outliers in any residual component, and based on the data distribution characteristics of outliers in any residual component, obtain the degree of concentration of outliers in any residual component; The sum of the deviation from the normal distribution and the concentration of outliers is obtained to get the decomposition anomaly degree of any residual component. The ratio of the decomposition anomaly degree to constant 4 is calculated, and the difference between constant 1 and the ratio is obtained to get the decomposition effect score of any residual component.
6. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 5, characterized in that, The step of obtaining the degree of deviation from the normal distribution of any residual component based on the difference in data distribution characteristics between any residual component and the standard normal distribution includes: Obtain the cumulative distribution function of any residual component and the standard normal distribution respectively, calculate the absolute value of the difference between the cumulative distribution function of any residual component and the cumulative distribution function of the standard normal distribution, obtain the absolute value function of the difference, and normalize the integral result of the absolute value function of the difference within its preset range to obtain the first normal distribution deviation degree; Obtain the absolute value of the difference between the maximum value of any residual component and the constant 1 to obtain the degree of deviation of the maximum value; calculate the ratio of the degree of deviation of the maximum value to the minimum value of any residual component to obtain the relative deviation value; obtain the difference between the constant 1 and the relative deviation value to obtain the degree of deviation of the second normal distribution. The degree of deviation from the normal distribution of any residual component is obtained by adding the first degree of deviation from the normal distribution to the second degree of deviation from the normal distribution.
7. The intelligent detection method for secondary circuit defects based on infrared thermometry according to claim 5, characterized in that, The step of obtaining the degree of outlier concentration in any residual component based on the data distribution characteristics of outliers in any residual component includes: Obtain the percentage of all outliers in any residual component to obtain the first outlier concentration. The outliers in any residual component are formed into an outlier sequence. The time interval between every two adjacent outliers in the outlier sequence is obtained, and the average time interval is obtained. The negative of the average time interval is substituted into an exponential function with the natural constant as the base to obtain the second outlier concentration. The degree of outlier concentration of any residual component is obtained by adding the first degree of outlier concentration to the second degree of outlier concentration.