Transformer oil dissolved gas early warning system and method based on hawkes process

Abstract

The application provides a transformer oil dissolved gas early warning system based on the Hough process, comprising: a historical data fitting module, which is used for dividing the historical concentration curve of a specific dissolved gas according to fluctuation periods, selecting the rising curve from the valley value to the peak value in each period, fitting the rising curves of all periods to obtain a unified fitting function; a Hough process modeling module, which is used for Hough process modeling according to the fitting functions of the rising curves of all fluctuation periods to obtain a conditional intensity function used for calculating the expected number of times of transformer failures corresponding to the dissolved gas; and a dissolved gas early warning module, which is used for calculating the expected number of times of transformer failures corresponding to the dissolved gas in a future time period according to the conditional intensity function of the expected number of times, and issuing an early warning in combination with an expected number of times threshold. The application shortens the time difference from failure to early warning, and overcomes the hysteresis problem caused by slow gas diffusion.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power transformers in electrical engineering, and specifically to an early warning system and method for dissolved gases in transformer oil based on the Hawkes process. Background Technology

[0002] Online monitoring of dissolved gases in transformer oil is one of the important means of condition monitoring for power transformers. Currently, the threshold comparison method is mainly used for fault early warning. However, this method has two significant limitations: first, the diffusion process of gases in oil is relatively slow, resulting in a significant lag in fault early warning; second, the dissolved gas concentration will decrease over time due to factors such as dissolution equilibrium, affecting the threshold judgment.

[0003] Current improved early warning methods mostly use gas concentration prediction. Traditional trend fitting methods only extrapolate the current trend, but the fault that caused the current trend change may have already occurred. They cannot predict the future faults and their probabilities. In addition, the prediction methods based on neural networks have insufficient interpretability, which is not conducive to fault diagnosis and decision-making in practical engineering applications. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a Hawkes process-based early warning system for dissolved gases in transformer oil. The system includes: a historical data fitting module, used to divide the historical fluctuation curve of the concentration of a certain dissolved gas in transformer oil over time into fluctuation periods. Within each fluctuation period, the system selects the rising curve of that fluctuation period starting from the valley value of the dissolved gas concentration and ending at the peak value of the dissolved gas concentration. The system then fits the rising curves of all fluctuation periods to obtain a fitting function for the rising curves of all fluctuation periods; a Hawkes process modeling module, used to perform Hawkes process modeling based on the fitting function of the rising curves of all fluctuation periods to obtain a conditional strength function for calculating the expected number of faults in the transformer corresponding to the dissolved gas; and a dissolved gas early warning module, used to calculate the expected number of faults in the transformer corresponding to the dissolved gas in the future time period based on the conditional strength function of the expected number of occurrences. The system compares the expected number of faults in the future time period with a set expected number threshold and determines whether to issue an early warning based on the comparison result.

[0005] Furthermore, the historical data fitting module is also used for:

[0006] Set a minimum peak-to-valley difference threshold, and denot the concentration range of the dissolved gas in the rising curves of all fluctuation periods where the peak-to-valley difference is greater than the minimum peak-to-valley difference threshold as [V]. i ,P i ], Vi P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of the fluctuation period i.

[0007] Furthermore, the historical data fitting module is also used for:

[0008] The Logistic function is used to fit the rising curves of all fluctuation cycles, as shown in the following formula:

[0009]

[0010] Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. i b is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c i It is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i;

[0011] Initialize the above parameters:

[0012] a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ;

[0013] Establish boundary constraints for the above parameters:

[0014] 0 i ≤2(P i -V i );0 i ≤1; 0.9×V i ≤d i ≤1.1×P i ;

[0015] The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function:​​

[0016]

[0017] Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

[0018] Furthermore, in the Hawkes process modeling module, the specific formula for the conditional strength function is as follows:

[0019]

[0020]

[0021] in, This is a conditional strength function used to calculate the expected number of dissolved gas faults in a transformer. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i;

[0022] The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β):

[0023] .

[0024] Furthermore, in the dissolved gas early warning module, the formula for the expected number of transformer faults corresponding to this type of dissolved gas occurring within a future time period is as follows:

[0025]

[0026] Where E[N(Δt)] is the expected number of faults to occur within the future time period [T, T+Δt], where T is a future time and Δt is the length of the time period;

[0027] When E[N(Δt)] < 1.5, continue monitoring and do not issue an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle and do not issue an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis.

[0028] Furthermore, the dissolved gas early warning module is also used to calculate the increase of this gas in the transformer oil during the future time period based on the expected number of transformer faults corresponding to this dissolved gas and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles. The formula for calculating the increase is as follows:

[0029] R = E[N(Δt)] × a max

[0030] Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

[0031] A method for early warning of dissolved gases in transformer oil based on the Hawkes process includes: dividing the historical fluctuation curve of the concentration of a certain dissolved gas in the transformer oil over time into fluctuation periods; selecting the rising curve of each fluctuation period starting from the valley value of the concentration of the dissolved gas and ending at the peak value of the concentration of the dissolved gas; fitting the rising curves of all fluctuation periods to obtain a fitting function for the rising curves of all fluctuation periods; modeling a Hawkes process based on the fitting function of the rising curves of all fluctuation periods to obtain a conditional strength function for calculating the expected number of faults corresponding to the dissolved gas in the transformer; calculating the expected number of faults corresponding to the dissolved gas in the transformer in the future time period based on the conditional strength function of the expected number of times; comparing the expected number of faults corresponding to the dissolved gas in the transformer in the future time period with a set expected number threshold; and determining whether to issue an early warning based on the comparison result.

[0032] Furthermore, the specific method for selecting the rising curve of each fluctuation cycle, starting from the valley value of the concentration of the dissolved gas and ending at the peak value of the concentration of the dissolved gas, is as follows:

[0033] Set a minimum peak-to-valley difference threshold, and denot the concentration range of the dissolved gas in the rising curves of all fluctuation periods where the peak-to-valley difference is greater than the minimum peak-to-valley difference threshold as [V]. i ,P i ], V i P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of the fluctuation period i.

[0034] Furthermore, the specific method for fitting the rising curves of all fluctuation cycles to obtain the fitting function for the rising curves of all fluctuation cycles is as follows:

[0035] The Logistic function is used to fit the rising curves of all fluctuation cycles, as shown in the following formula:

[0036]

[0037] Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. i b is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c i It is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i;

[0038] Initialize the above parameters:

[0039] a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ;

[0040] Establish boundary constraints for the above parameters:

[0041] 0 i ≤2(P i -V i );0 i ≤1; 0.9×V i ≤d i ≤1.1×P i ;

[0042] The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function:

[0043]

[0044] Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k ​​It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

[0045] Furthermore, the specific method for modeling the Hawkes process based on the fitting function of the rising curves of all fluctuation cycles to obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this dissolved gas is as follows:

[0046]

[0047]

[0048] in, This is a conditional strength function used to calculate the expected number of faults in a transformer corresponding to a given dissolved gas. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i;

[0049] The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β):

[0050] .

[0051] Furthermore, the specific method for calculating the expected number of transformer faults corresponding to the dissolved gas in the future time period based on the conditional strength function of the expected number of occurrences, comparing the expected number of transformer faults corresponding to the dissolved gas in the future time period with a set expected number threshold, and determining whether to issue a warning based on the comparison result is as follows:

[0052]

[0053] Where E[N(Δt)] is the expected number of faults to occur within the future time period [T, T+Δt], where T is a future time and Δt is the length of the time period;

[0054] When E[N(Δt)] < 1.5, continue monitoring and do not issue an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle and do not issue an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis.

[0055] Furthermore, based on the expected number of transformer faults corresponding to this dissolved gas occurring within the future time period and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles, the increase in this gas in the transformer oil within the future time period is calculated. The specific method is as follows:

[0056] R = E[N(Δt)] × a max

[0057] Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

[0058] A computer-readable medium storing a computer program / instructions that, when executed, perform the aforementioned method for early warning of dissolved gases in transformer oil based on the Hawkes process.

[0059] An electronic device is characterized in that it comprises: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer programs / instructions, and the processor executing the computer programs / instructions to perform the aforementioned method for early warning of dissolved gases in transformer oil based on the Hawkes process.

[0060] The beneficial effects of this invention are as follows:

[0061] 1. This invention models the self-excitation characteristics of historical fault events through Hawkes processes and dynamically calculates the expected number of faults E[N(Δt) in the future time period using the conditional intensity function λ(t). Compared with the traditional threshold comparison method, which relies on the passive early warning of current concentration exceeding the standard, this method can predict the probability of future faults in advance, significantly shortening the time difference from fault occurrence to early warning and overcoming the hysteresis problem caused by slow gas diffusion.

[0062] 2. By segmenting the rising curves in historical data, starting from the concentration trough and ending at the peak, and accurately fitting the curves using the Logistic function, this method focuses only on the concentration rise phase caused by the fault. This effectively isolates the natural decreasing trend of the gas due to dissolution equilibrium and solves the risk of misjudgment caused by concentration fluctuations in traditional threshold methods.

[0063] 3. A Hawkes process model is adopted to replace the neural network black-box model. Parameters are solved by maximizing the likelihood function, making the prediction process transparent and traceable. Simultaneously, the expected number of faults E[N(Δt)] and the maximum concentration difference a are combined. max It calculates the future gas growth rate R, providing dual quantitative basis for "whether to issue an early warning" and "changing the monitoring cycle", supporting precise operation and maintenance decisions. Attached Figure Description

[0064] Figure 1 This is a system block diagram of the present invention.

[0065] Figure 2 This is a flowchart of the method of the present invention.

[0066] Figure 3 This is a historical fluctuation curve of the C2H2 concentration over the past 50 days for a certain transformer.

[0067] Figure 4 This is a schematic diagram of the segmentation points after data smoothing and data segmentation of the historical fluctuation curve of C2H2 concentration over time.

[0068] Figure 5 This is a schematic diagram after fitting the rising curve. Detailed Implementation

[0069] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0070] This invention is based on Hawkes process theory. By modeling the time-series correlation of historical gas-producing events, it achieves prediction and early warning of future gas-producing faults. The Hawkes process is a point process model with self-excitation characteristics, capable of capturing triggering effects and clustering phenomena between events. Its core feature is that historical events enhance the intensity of future events; this influence decays exponentially over time and then accumulates. Related transformer fault cases show that the progression from sporadic gas-producing events to high-frequency and high-concentration gas production, culminating in severe faults, also exhibits triggering effects and clustering phenomena.

[0071] Example 1

[0072] like Figure 1 The aforementioned, a dissolved gas early warning system for transformer oil based on the Hawkes process, includes:

[0073] The historical data fitting module is used to divide the historical fluctuation curve of the concentration of a certain dissolved gas in transformer oil over time into fluctuation periods. In each fluctuation period, the rising curve of the fluctuation period is selected with the valley value of the concentration of the dissolved gas as the starting point and the peak value of the concentration of the dissolved gas as the ending point. The rising curves of all fluctuation periods are fitted to obtain the fitting function of the rising curves of all fluctuation periods.

[0074] The Hawkes process modeling module is used to model the Hawkes process based on the fitting function of the rising curve of all fluctuation cycles, and obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this dissolved gas.

[0075] The dissolved gas early warning module is used to calculate the expected number of transformer faults corresponding to a certain dissolved gas in the future time period based on the conditional strength function of the expected number, compare the expected number of transformer faults corresponding to a certain dissolved gas in the future time period with a set expected number threshold, and determine whether to issue an early warning based on the comparison result.

[0076] (1) The historical data fitting module is also used for:

[0077] Collect historical fluctuation curves of the concentration of dissolved gases such as H2, CH4, and C2H2 in transformer oil over time (historical data within the most recent month, such as...). Figure 3 As shown in the figure, the data is preprocessed, such as removing values ​​with a deviation exceeding 3 times the variance within a certain time range, and filling in missing values ​​using interpolation. The data for each component can be processed separately; for example, C2H2 is mainly used to focus on discharge faults, CH4 is mainly used to focus on overheating faults, and H2 is used to help determine its correlation with other gas production.

[0078] The Savitzky-Golay filter (SG filter) is used to eliminate high-frequency noise in the data. It is a smoothing filtering method based on local polynomial least squares fitting. Its core idea is to perform polynomial fitting on the data within a sliding window and replace the original data with the fitted values, thereby smoothing noise while preserving as many high-order features of the signal as possible (such as peaks and inflection points). For a discrete-time series x={x1,x2,…,x…}, the following method is applied. N}, its expression is:

[0079]

[0080] Among them, y t Here is the filtered gas concentration data, where t is time, K is the half-window width (total window width W = 2K + 1), and c k The polynomial coefficients are determined for least-squares fitting. Recommended parameters: 11-point window width, 3rd-order polynomial. Since the SG filter cannot be computed at the data boundaries (first K points and last K points), a smaller window (e.g., linear extrapolation) is used for the boundary points.

[0081] For the historical fluctuation curve of the concentration of the filtered dissolved gas over time, a minimum peak-to-valley difference threshold is set. The concentration range of this dissolved gas in the rising curve of all fluctuation periods with a peak-to-valley difference greater than the minimum peak-to-valley difference threshold is denoted as [V]. i ,P i ], Vi P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of fluctuation period i. Here, fluctuation period refers to a segment of the historical fluctuation curve that shows a curve that first rises and then falls or first rises and then flattens out. The historical fluctuation curve includes multiple fluctuation periods, and each fluctuation period corresponds to one transformer fault.

[0082] Historical time-series data of dissolved gas concentrations in the oil of a transformer were collected over the past 50 days. Taking C2H2 as an example, after removing obvious outliers and completing the data, the data trend is as follows: Figure 3 As shown, the horizontal axis is in h and the vertical axis is in ppm. The data shows an increasing trend, stabilizing around 1.0 ppm. If a conventional trend-fitting prediction method is used, the transformer fault development might be considered to be trending towards stability. Local peaks and troughs are detected in the filtered data. The data is segmented according to peaks and troughs, and a minimum peak-to-trough difference threshold is set (e.g., 5% of the concentration baseline). The minimum distance between adjacent peaks and troughs is set to 3 to avoid noise interference. Figure 4 As shown, peak points are marked with "×" and valley points are marked with "●". After segmentation, the data is divided into rising segments and falling segments.

[0083] Traditional methods are sensitive to small fluctuations in gas concentration and are prone to misjudgment due to noise or natural fluctuations (as mentioned in the background art, "the downward trend caused by dissolved gas balance affects threshold judgment"). This invention reduces the false alarm rate by eliminating small fluctuations (such as <5% of the baseline value) through thresholding, and only retaining the significant concentration increase segment caused by real faults.

[0084] (2) The historical data fitting module is also used for:

[0085] The rising curves of all fluctuation cycles were fitted using the Logistic function. The Logistic function (also known as the Sigmoid function) is a classic S-shaped growth curve. The changes in dissolved gas concentration in oil typically exhibit a trend of "slow growth - rapid rise - saturation and stabilization," which highly matches the S-shaped growth characteristics of the Logistic function. The formula is as follows:

[0086]

[0087] Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. i b is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c iIt is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i;

[0088] Initialize the above parameters:

[0089] a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ;

[0090] Establish boundary constraints for the above parameters:

[0091] 0 i ≤2(P i -V i );0 i ≤1; 0.9×V i ≤d i ≤1.1×P i ;

[0092] The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function:

[0093]

[0094] Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

[0095] After fitting the rising curve, as follows Figure 5 As shown.

[0096] Traditional linear / polynomial fitting cannot capture the "S-shaped growth" characteristic of fault gas release (slow start → accelerated rise → saturation and stabilization). This invention uses the Logistic function to naturally match the nonlinear growth process of fault gas production, with clearly defined physical meanings for its parameters (such as b).​​i (Directly reflects the speed of fault development), initialization strategies accelerate the optimization process, and boundary constraints prevent the fitting results from deviating from physical reality. The Levenberg-Marquardt algorithm has strong noise resistance, ensuring reliable fitting results.

[0097] (3) In the Hawkes process modeling module, the specific formula for the conditional strength function is as follows:

[0098]

[0099]

[0100] in, This is a conditional strength function used to calculate the expected number of faults in a transformer corresponding to a given dissolved gas. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i;

[0101] The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β):

[0102] .

[0103] The starting point t of each ascending segment Vi This is marked as a potential gas production failure event. From Figure 4 As can be seen above, there are a total of 9 rising segments, and the marked gas production event time points are {0, 156, 316, 436, 600, 744, 880, 968, 1016}, in hours.

[0104] The fitted parameters and calculated weights are shown in Table 1.

[0105] Table 1

[0106]

[0107] When finding the optimal values ​​of μ, α, and β by maximizing the likelihood function L(μ,α,β), the EM algorithm or gradient descent method is used. The total duration T is 1196 hours. Based on the data in the table above, the fitting results are μ=0.01, α=0.50, and β=0.01.

[0108] Existing neural network methods lack interpretability and cannot quantify the triggering effect of historical faults on future faults. This invention addresses this by constructing a Hawkes process... ifusion concentration change range a i and speed b i This comprehensively characterizes the severity of historical faults. The exponentially decaying term e -β(t-tk) Accurate modeling of the timeliness of fault impacts. μ (inherent gas production rate), α (fault self-excitation intensity), and β (impact decay rate) have clear physical meanings, supporting root cause analysis of faults. Maximizing the likelihood function ensures the statistical optimality of parameter estimation.

[0109] (4) In the dissolved gas early warning module, the formula for the expected number of transformer faults corresponding to the dissolved gas in the future time period is as follows:

[0110]

[0111] Where E[N(Δt)] is the expected number of faults to occur within the future time period [T,T+Δt], where T is a future time and Δt is the length of the time period.

[0112] When E[N(Δt)] < 1.5, continue monitoring without issuing an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle without issuing an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis. Further judgment is then made based on whether R exceeds the threshold specified in the relevant standards.

[0113] Based on the expected number of transformer faults corresponding to this dissolved gas occurring within the future time period and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles, the increase in this gas in the transformer oil within the future time period is calculated. The formula for calculating the increase is as follows:

[0114] R = E[N(Δt)] × a max

[0115] Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

[0116] Based on the above data, the predicted number of events in the next 168 hours is E[N(Δt)] = 2.27. When E[N(Δt)] ≥ 2, an early warning should be issued to initiate offline detection and diagnosis. The maximum value of the fault amplitude 'a' over the past 50 days is also considered. max =0.402, R=2.27×0.402=0.91, predicting that the acetylene content will increase by about 0.91 ppm in the next 168 hours.

[0117] Example 2

[0118] like Figure 2As shown, a method for early warning of dissolved gases in transformer oil based on the Hawkes process includes:

[0119] The historical fluctuation curve of the concentration of a certain dissolved gas in transformer oil over time is divided into fluctuation periods. In each fluctuation period, the rising curve of the fluctuation period is selected with the valley value of the concentration of the dissolved gas as the starting point and the peak value of the concentration of the dissolved gas as the ending point. The rising curves of all fluctuation periods are fitted to obtain the fitting function of the rising curves of all fluctuation periods.

[0120] Hawkes process modeling is performed based on the fitting function of the rising curves of all fluctuation cycles to obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this dissolved gas.

[0121] Based on the conditional strength function of the expected number of occurrences, the expected number of transformer faults corresponding to the dissolved gas in the future time period is calculated. The expected number of transformer faults corresponding to the dissolved gas in the future time period is compared with the set expected number threshold, and a warning is issued based on the comparison result.

[0122] (1) The specific method for selecting the rising curve of each fluctuation cycle, starting from the valley value of the concentration of the dissolved gas and ending at the peak value of the concentration of the dissolved gas, is as follows:

[0123] Collect historical fluctuation curves of the concentration of dissolved gases such as H2, CH4, and C2H2 in transformer oil over time (historical data within the most recent month, such as...). Figure 3 As shown in the figure, the data is preprocessed, such as removing values ​​with a deviation exceeding 3 times the variance within a certain time range, and filling in missing values ​​using interpolation. The data for each component can be processed separately; for example, C2H2 is mainly used to focus on discharge faults, CH4 is mainly used to focus on overheating faults, and H2 is used to help determine its correlation with other gas production.

[0124] The Savitzky-Golay filter (SG filter) is used to eliminate high-frequency noise in the data. It is a smoothing filtering method based on local polynomial least squares fitting. Its core idea is to perform polynomial fitting on the data within a sliding window and replace the original data with the fitted values, thereby smoothing noise while preserving as many high-order features of the signal as possible (such as peaks and inflection points). For a discrete-time series x={x1,x2,…,x…}, the following method is applied. N}, its expression is:

[0125]

[0126] Among them, y tHere is the filtered gas concentration data, where t is time, K is the half-window width (total window width W = 2K + 1), and c k The polynomial coefficients are determined for least-squares fitting. Recommended parameters: 11-point window width, 3rd-order polynomial. Since the SG filter cannot be computed at the data boundaries (first K points and last K points), a smaller window (e.g., linear extrapolation) is used for the boundary points.

[0127] For the historical fluctuation curve of the concentration of the filtered dissolved gas over time, a minimum peak-to-valley difference threshold is set. The concentration range of this dissolved gas in the rising curve of all fluctuation periods with a peak-to-valley difference greater than the minimum peak-to-valley difference threshold is denoted as [V]. i ,P i ], V i P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of the fluctuation period i.

[0128] Historical time-series data of dissolved gas concentrations in the oil of a transformer were collected over the past 50 days. Taking C2H2 as an example, after removing obvious outliers and completing the data, the data trend is as follows: Figure 3 As shown, the horizontal axis is in h and the vertical axis is in ppm. The data shows an increasing trend, stabilizing around 1.0 ppm. If a conventional trend-fitting prediction method is used, the transformer fault development might be considered to be trending towards stability. Local peaks and troughs are detected in the filtered data. The data is segmented according to peaks and troughs, and a minimum peak-to-trough difference threshold is set (e.g., 5% of the concentration baseline). The minimum distance between adjacent peaks and troughs is set to 3 to avoid noise interference. Figure 4 As shown, peak points are marked with "×" and valley points are marked with "●". After segmentation, the data is divided into rising segments and falling segments.

[0129] (2) The specific method for fitting the rising curves of all fluctuation cycles to obtain the fitting function of the rising curves of all fluctuation cycles is as follows:

[0130] The rising curves of all fluctuation cycles were fitted using the Logistic function. The Logistic function (also known as the Sigmoid function) is a classic S-shaped growth curve. The changes in dissolved gas concentration in oil typically exhibit a trend of "slow growth - rapid rise - saturation and stabilization," which highly matches the S-shaped growth characteristics of the Logistic function. The formula is as follows:

[0131]

[0132] Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. ib is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c i It is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i;

[0133] Initialize the above parameters:

[0134] a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ;

[0135] Establish boundary constraints for the above parameters:

[0136] 0 i ≤2(P i -V i );0 i ≤1; 0.9×V i ≤d i ≤1.1×P i ;

[0137] The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function:

[0138]

[0139] Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

[0140] After fitting the rising curve, as follows Figure 5 As shown.

[0141] ​​(3) The specific method for modeling the Hawkes process based on the fitting function of the rising curves of all fluctuation cycles to obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this type of dissolved gas is as follows:

[0142]

[0143]

[0144] in, This is a conditional strength function used to calculate the expected number of faults in a transformer corresponding to a given dissolved gas. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i;

[0145] The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β):

[0146] .

[0147] The starting point t of each ascending segment Vi This is marked as a potential gas production failure event. From Figure 4 As can be seen above, there are a total of 9 rising segments, and the marked gas production event time points are {0, 156, 316, 436, 600, 744, 880, 968, 1016}, in hours.

[0148] The fitted parameters and calculated weights are shown in Table 1.

[0149] When finding the optimal values ​​of μ, α, and β by maximizing the likelihood function L(μ,α,β), the EM algorithm or gradient descent method is used. The total duration T is 1196 hours. Based on the data in the table above, the fitting results are μ=0.01, α=0.50, and β=0.01.

[0150] (4) The specific method for calculating the expected number of transformer faults corresponding to the dissolved gas in the future time period based on the conditional strength function of the expected number of times, comparing the expected number of transformer faults corresponding to the dissolved gas in the future time period with the set expected number threshold, and determining whether to issue a warning based on the comparison result is as follows:

[0151]

[0152] Where E[N(Δt)] is the expected number of faults to occur within the future time period [T,T+Δt], where T is a future time and Δt is the length of the time period.

[0153] When E[N(Δt)] < 1.5, continue monitoring without issuing an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle without issuing an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis. Further judgment is then made based on whether R exceeds the threshold specified in the relevant standards.

[0154] The specific method for calculating the increase of this gas in transformer oil during the future time period is as follows: Based on the expected number of transformer faults corresponding to this dissolved gas occurring within the future time period and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles, the method is as follows:

[0155] R = E[N(Δt)] × a max

[0156] Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

[0157] Based on the above data, the predicted number of events in the next 168 hours is E[N(Δt)] = 2.27. When E[N(Δt)] ≥ 2, an early warning should be issued to initiate offline detection and diagnosis. The maximum value of the fault amplitude 'a' over the past 50 days is also considered. max =0.402, R=2.27×0.402=0.91, predicting that the acetylene content will increase by about 0.91 ppm in the next 168 hours.

[0158] Example 3

[0159] A computer-readable medium storing a computer program that, when executed, performs the method for early warning of dissolved gases in transformer oil based on the Hawkes process in Embodiment 2.

[0160] Example 4

[0161] An electronic device is characterized in that it comprises: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method for early warning of dissolved gases in transformer oil based on the Hawkes process in Embodiment 2.

[0162] The contents not described in detail in this specification are prior art known to those skilled in the art. Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0163] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0164] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0165] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A dissolved gas early warning system for transformer oil based on the Hawkes process, characterized in that, include: The historical data fitting module is used to divide the historical fluctuation curve of the concentration of a certain dissolved gas in transformer oil over time into fluctuation periods. In each fluctuation period, the rising curve of the fluctuation period is selected with the valley value of the concentration of the dissolved gas as the starting point and the peak value of the concentration of the dissolved gas as the ending point. The rising curves of all fluctuation periods are fitted to obtain the fitting function of the rising curves of all fluctuation periods. The Hawkes process modeling module is used to model the Hawkes process based on the fitting function of the rising curve of all fluctuation cycles, and obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this dissolved gas. The dissolved gas early warning module is used to calculate the expected number of transformer faults corresponding to the dissolved gas in the future time period based on the conditional strength function of the expected number, compare the expected number of transformer faults corresponding to the dissolved gas in the future time period with the set expected number threshold, and determine whether to issue an early warning based on the comparison result. In the Hawkes process modeling module, the specific formula for the conditional strength function is as follows: in, This is a conditional strength function used to calculate the expected number of faults in a transformer corresponding to a given dissolved gas. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i, where t is time. k It is the time corresponding to sampling point k in the rising curve of fluctuation period i; The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β): ； In the dissolved gas early warning module, the formula for the expected number of transformer faults corresponding to a certain dissolved gas within a future time period is as follows: Where E[N(Δt)] is the expected number of faults to occur within the future time period [T, T+Δt], where T is a future time and Δt is the length of the time period; When E[N(Δt)] < 1.5, continue monitoring and do not issue an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle and do not issue an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis.

2. The early warning system for dissolved gases in transformer oil based on the Hawkes process according to claim 1, characterized in that: The historical data fitting module is also used for: Set a minimum peak-to-valley difference threshold, and denot the concentration range of the dissolved gas in the rising curves of all fluctuation periods where the peak-to-valley difference is greater than the minimum peak-to-valley difference threshold as [V]. i ,P i ], V i P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of the fluctuation period i.

3. The early warning system for dissolved gases in transformer oil based on the Hawkes process according to claim 2, characterized in that: The historical data fitting module is also used for: The Logistic function is used to fit the rising curves of all fluctuation cycles, as shown in the following formula: Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. i b is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c i It is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i; Initialize the above parameters: a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ; Establish boundary constraints for the above parameters: 0 i ≤2(P i -V i );0 i ≤1;0.9×V i ≤d i ≤1.1×P i ;​​ The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function: Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

4. The early warning system for dissolved gases in transformer oil based on the Hawkes process according to claim 1, characterized in that: The dissolved gas early warning module is also used to calculate the increase of this gas in the transformer oil during the future time period based on the expected number of transformer faults corresponding to this dissolved gas and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles. The formula for calculating the increase is as follows: R=E[N(Δt)]×a max Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

5. A method for early warning of dissolved gases in transformer oil based on the Hawkes process, characterized in that, include: The historical fluctuation curve of the concentration of a certain dissolved gas in transformer oil over time is divided into fluctuation periods. In each fluctuation period, the rising curve of the fluctuation period is selected with the valley value of the concentration of the dissolved gas as the starting point and the peak value of the concentration of the dissolved gas as the ending point. The rising curves of all fluctuation periods are fitted to obtain the fitting function of the rising curves of all fluctuation periods. Hawkes process modeling is performed based on the fitting function of the rising curves of all fluctuation cycles to obtain the conditional strength function for calculating the expected number of transformer faults corresponding to this dissolved gas. Based on the conditional strength function of the expected number of occurrences, the expected number of transformer faults corresponding to the dissolved gas in the future time period is calculated. This expected number of faults is then compared with a set threshold for the expected number of occurrences, and a warning is issued based on the comparison result. The specific formula for the conditional strength function in the Hawkes process modeling module is as follows: in, This is a conditional strength function used to calculate the expected number of faults in a transformer corresponding to a given dissolved gas. μ is the inherent gas production rate of the transformer under conditions of natural aging or normal operation without external fault stimuli. α is the fault self-excitation coefficient, which represents the instantaneous enhancement of the current fault occurrence rate by a single historical fault. β is the fault attenuation coefficient, which controls the attenuation rate of the influence of historical faults. i It is the weight of the rising curve of fluctuation period i, where t is time. k It is the time corresponding to sampling point k in the rising curve of fluctuation period i; The optimal values ​​of μ, α, and β are found by maximizing the likelihood function L(μ,α,β): ； In the dissolved gas early warning module, the formula for the expected number of transformer faults corresponding to a certain dissolved gas within a future time period is as follows: Where E[N(Δt)] is the expected number of faults to occur within the future time period [T, T+Δt], where T is a future time and Δt is the length of the time period; When E[N(Δt)] < 1.5, continue monitoring and do not issue an early warning; when 1.5 ≤ E[N(Δt)] < 2, shorten the monitoring cycle and do not issue an early warning; when E[N(Δt)] ≥ 2, issue an early warning to initiate offline detection and diagnosis.

6. The method for early warning of dissolved gases in transformer oil based on the Hawkes process according to claim 5, characterized in that: The specific method for selecting the rising curve of each fluctuation cycle, starting from the valley value of the concentration of the dissolved gas and ending at the peak value of the concentration of the dissolved gas, is as follows: Set a minimum peak-to-valley difference threshold, and denot the concentration range of the dissolved gas in the rising curves of all fluctuation periods where the peak-to-valley difference is greater than the minimum peak-to-valley difference threshold as [V]. i ,P i ], V i P is the valley value of the dissolved gas concentration in the rising curve of the fluctuation period i. i It is the peak concentration of the dissolved gas in the rising curve of the fluctuation period i.

7. The method for early warning of dissolved gases in transformer oil based on the Hawkes process according to claim 6, characterized in that: The specific method for fitting the rising curves of all fluctuation cycles to obtain the fitting function for the rising curves of all fluctuation cycles is as follows: The Logistic function is used to fit the rising curves of all fluctuation cycles, as shown in the following formula: Among them, f i (t) is the fitting function for the rising curve of fluctuation period i, where t is time and a is the value of time. i b is the peak-to-valley difference in the concentration of this dissolved gas in the rising curve of fluctuation period i. i It is the rate of increase of the concentration of this dissolved gas in the rising curve of the fluctuation period i, c i It is the point in time where the concentration of this dissolved gas increases at the fastest rate in the rising curve of fluctuation period i, d i It is the initial dissolved concentration of the gas in the rising curve of the fluctuation period i; Initialize the above parameters: a i =P i -V i b i =(P i -V i ) / t;c i Set as [V] i ,P i The midpoint of ]; d i =V i ; Establish boundary constraints for the above parameters: 0 i ≤2(P i -V i );0 i ≤1;0.9×V i ≤d i ≤1.1×P i ;​​ The Levenberg-Marquardt algorithm is used to solve for a. i b i c i d i The optimal value is determined using the following objective function: Where k is the sampling point number of the time series in the rising curve of fluctuation period i, and t k The sampling point k in the rising curve of fluctuation period i corresponds to the time x. i,k It is the rising curve of the fluctuation period i at time t k The dissolved gas concentration is N, and N is the total number of sampling points in the time series of the rising curve of the fluctuation period i.

8. The method for early warning of dissolved gases in transformer oil based on the Hawkes process according to claim 5, characterized in that: Based on the expected number of transformer faults corresponding to this dissolved gas occurring within the future time period and the maximum concentration difference of this dissolved gas in the rising curves of all fluctuation cycles, the increase of this gas in the transformer oil within the future time period is calculated. The specific method is as follows: R=E[N(Δt)]×a max Where R represents the increase in the amount of this gas in the transformer oil during the future time period, and a max It is the maximum concentration difference of this dissolved gas in the rising curve of all fluctuation cycles.

9. A computer-readable medium, characterized in that, The computer-readable medium stores a computer program / instruction that, when executed, performs the Hawkes process-based method for early warning of dissolved gases in transformer oil as described in any one of claims 5-8.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer programs / instructions, and the processor executing the computer programs / instructions to perform the method for early warning of dissolved gases in transformer oil based on the Hawkes process as described in any one of claims 5-8.

Images