A method, device and equipment for predicting operation risk of oil-immersed transformer and storage medium

Abstract

This invention discloses a method, apparatus, equipment, and storage medium for predicting the operational risks of oil-immersed transformers. The method involves acquiring oil chromatography data sequences and operating environment data sequences, and preprocessing these sequences. A Gram angle difference field matrix is ​​constructed based on the preprocessed oil chromatography data sequences, and the Gram angle difference field matrix is ​​stacked to generate a multi-channel feature map sequence. This multi-channel feature map sequence is input into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence. A second risk value sequence is determined based on the first risk value sequence and the operating environment data sequence. A gated recurrent unit network is used to predict risks based on the first global feature sequence and the operating environment data sequence to obtain at least one second predicted risk value. Curve fitting is performed on the second risk value sequence and at least one second predicted risk value to obtain a risk curve. This method can improve the accuracy of risk prediction.

Description

Technical Field

[0001] This invention relates to the field of oil-immersed transformer technology, and in particular to a method, apparatus, equipment and storage medium for predicting operational risks of oil-immersed transformers. Background Technology

[0002] In recent years, with the development of sensor technology, online monitoring technology, and data acquisition systems, multi-source heterogeneous monitoring data has gradually achieved real-time acquisition, providing a foundation for data-driven intelligent diagnostic methods. Against this backdrop, transformer condition assessment methods combining machine learning and deep learning technologies have gradually emerged, improving diagnostic accuracy to some extent by exploring the nonlinear mapping relationship between oil chromatography data and fault types. However, existing technologies still have several shortcomings in practical applications. Summary of the Invention

[0003] This invention provides a method, apparatus, equipment, and storage medium for predicting the operational risks of oil-immersed transformers, which can improve the accuracy of predicting the operational risks of oil-immersed transformers.

[0004] In a first aspect, embodiments of the present invention provide a method for predicting operational risks of oil-immersed transformers, comprising:

[0005] Obtain oil chromatography data sequences and operating environment data sequences, and preprocess the oil chromatography data sequences and the operating environment data sequences;

[0006] A Gram angle difference field matrix is ​​constructed based on the preprocessed oil chromatography data sequence, and the Gram angle difference field matrix is ​​stacked to generate a multi-channel feature map sequence.

[0007] The multi-channel feature map sequence is input into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence.

[0008] A second risk value sequence is determined based on the first risk value sequence and the operating environment data sequence;

[0009] Based on the first global feature sequence and the operating environment data sequence, a gated recurrent unit network is used to perform risk prediction to obtain at least one second predicted risk value.

[0010] A risk curve is obtained by curve fitting the second risk value sequence and the at least one second predicted risk value; wherein the risk curve is used to characterize the risk change trend of the oil-immersed transformer.

[0011] Secondly, embodiments of the present invention provide an operational risk prediction device for oil-immersed transformers, comprising:

[0012] The data acquisition module is used to acquire oil chromatography data sequences and operating environment data sequences, and to preprocess the oil chromatography data sequences and the operating environment data sequences.

[0013] The multi-channel feature map sequence generation module is used to construct a Gram angle difference field matrix based on the preprocessed oil chromatography data sequence, and to stack the Gram angle difference field matrix to generate a multi-channel feature map sequence.

[0014] The module for obtaining the first global feature sequence and the first risk value sequence is used to input the multi-channel feature map sequence into a convolutional neural network to obtain the first global feature sequence and the first risk value sequence.

[0015] The second risk value sequence determination module is used to determine a second risk value sequence based on the first risk value sequence and the operating environment data sequence.

[0016] The second predicted risk value determination module is used to perform risk prediction based on the first global feature sequence and the operating environment data sequence using a gated recurrent unit network to obtain at least one second predicted risk value.

[0017] The risk curve acquisition module is used to perform curve fitting on the second risk value sequence and the at least one second predicted risk value to obtain a risk curve; wherein the risk curve is used to characterize the risk change trend of the oil-immersed transformer.

[0018] Thirdly, embodiments of the present invention provide an electronic device, the electronic device comprising:

[0019] At least one processor; and a memory communicatively connected to said at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the operation risk prediction method for oil-immersed transformers according to any embodiment of the present invention.

[0021] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions, which are used to cause a processor to execute the method for predicting the operational risk of an oil-immersed transformer as described in any embodiment of the present invention.

[0022] This invention discloses a method, apparatus, device, and storage medium for predicting the operational risks of oil-immersed transformers. The method involves acquiring oil chromatography data sequences and operational environment data sequences, and preprocessing these sequences. A Gram angle difference field matrix is ​​constructed based on the preprocessed oil chromatography data sequences, and the Gram angle difference field matrix is ​​stacked to generate a multi-channel feature map sequence. This multi-channel feature map sequence is input into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence. A second risk value sequence is determined based on the first risk value sequence and the operational environment data sequence. Risk prediction is performed using a gated recurrent unit network based on the first global feature sequence and the operational environment data sequence to obtain at least one second predicted risk value. Curve fitting is performed on the second risk value sequence and at least one second predicted risk value to obtain a risk curve. The risk curve characterizes the risk change trend of the oil-immersed transformer. The operational risk prediction method for oil-immersed transformers provided in this invention constructs a Gram angle difference field matrix to transform oil chromatography data into a feature map suitable for convolutional neural network processing, thus preserving temporal dependencies; it determines a second risk value sequence using operating environment data and a first risk value sequence, improving the accuracy of risk value determination; and it performs risk prediction based on a first global feature sequence and operating environment data to predict risk values ​​at multiple future times, thereby improving prediction accuracy.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of a method for predicting operational risks of an oil-immersed transformer according to an embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of the structure of an operation risk prediction device for an oil-immersed transformer according to an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention 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 the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0031] Existing technologies still have several shortcomings in practical applications. First, traditional oil chromatography analysis methods treat time-series data as independent sampling points, ignoring the time dependence and evolution trajectory information of characteristic gas concentration changes, making it difficult to capture the gradual characteristics in the fault incubation process. Second, existing assessment methods are mostly limited to judging the current state and lack the ability to dynamically predict the risk evolution trend, making it difficult to provide forward-looking guidance for operation and maintenance decisions. Finally, when processing operating environment data such as partial discharge and temperature, existing methods often use simple threshold superposition or weighted summation, failing to fully consider the risk amplification effect under the coupling of multiple factors.

[0032] Example 1

[0033] Figure 1 This is a flowchart illustrating a method for predicting the operational risks of oil-immersed transformers according to an embodiment of the present invention. This embodiment is applicable to situations where the operational risks of oil-immersed transformers need to be predicted. The method can be executed by the operational risk prediction device for oil-immersed transformers described in this embodiment. This device can be implemented using software and / or hardware, such as... Figure 1 As shown, the method specifically includes the following steps:

[0034] S110: Acquire oil chromatography data sequences and operating environment data sequences, and preprocess the oil chromatography data sequences and operating environment data sequences.

[0035] The oil chromatographic data sequence may include the following gas concentration sequences dissolved in the oil: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide. The oil chromatographic data sequence can be obtained by periodically (e.g., every hour) collecting data on each gas concentration using sensors, and this data sequence may include data from a recent period, such as the last six months. The operating environment data sequence may include winding hotspot temperature sequences and partial discharge amplitude sequences, which can be obtained by periodically (e.g., every hour) collecting data on winding hotspot temperature and partial discharge amplitude using sensors, and this data sequence may include data from a recent period, such as the last six months.

[0036] Preprocessing may include timeline alignment, outlier correction, and normalization.

[0037] Specifically, the time axis alignment can be achieved by constructing a unified periodic (e.g., hourly) time axis, using ordinal spline interpolation to obtain initial estimates of each gas from the oil chromatographic data, and then correcting them by combining the non-negative gas growth rate constraint and the concentration smoothing constraint of adjacent time periods to obtain the reconstructed concentration of each gas on the unified hourly time axis. For the winding hot spot temperature, Kalman filtering with constraints introduced by the transformer thermal balance equation is used for state estimation to obtain the optimal estimate of the winding hot spot temperature on the unified hourly time axis. For the partial discharge amplitude data, the 95th percentile value of the partial discharge is calculated within each hourly window as the representative partial discharge amplitude for that hour.

[0038] Specifically, outlier correction can be achieved by constructing a sliding window W of a preset length (e.g., 24 hours) for any given time, establishing an isolated forest model within the sliding window W to detect outliers, replacing oil chromatographic outliers with order-preserving interpolation correction values ​​from adjacent time periods, replacing temperature outliers with Kalman filter estimates, and replacing partial discharge outliers with 95% quantile smoothed values ​​from the nearest hour window.

[0039] Specifically, the normalization method can be: perform minimum-maximum normalization on each gas concentration sequence, mapping the sequence values ​​to the interval [0,1].

[0040] S120: Based on the preprocessed oil chromatography data sequence, construct the Gram angle difference field matrix, and stack the Gram angle difference field matrix to generate a multi-channel feature map sequence.

[0041] The number of channels is related to the type of gas. In this embodiment, there are 7 types of oil-dissolving gases, so the number of channels is 7.

[0042] Optionally, the method for constructing a Gram angle difference field matrix based on the preprocessed oil chromatography data sequence and stacking the Gram angle difference field matrix to generate a multi-channel feature map sequence can be as follows: map the preprocessed gas concentration sequence to polar coordinate space to obtain an angle sequence; for each data point in the angle sequence, obtain the angle sub-sequence within a preset historical time period corresponding to the data point; calculate the sine value of the pairwise angle difference in the angle sub-sequence to obtain the Gram angle difference field matrix corresponding to each gas; stack the Gram angle difference field matrices corresponding to each gas to obtain the multi-channel feature map corresponding to the data point.

[0043] Among them, the multi-channel feature maps corresponding to each data point constitute a multi-channel feature map sequence.

[0044] The polar coordinates include the radius r and the angle. Of the two coordinates, only the angle information will be used in this embodiment. The formula for mapping the preprocessed gas concentration sequence to polar coordinate space can be expressed as: , ;in, The radius represents the position of time point j in the sequence, where j represents the index of the time point (also known as the data point), and N represents the sequence length. The angle indicates the numerical value of the gas concentration. Represents the normalized gas At the point of time The concentration value, where k represents the gas type index.

[0045] The Gram angle difference field matrix can be expressed as: ; ;in, Denotes the Gram angle difference field matrix of gas k. Let represent the sine of the angle difference between time point a and time point b, located in the a-th row and b-th column of the matrix. and These represent the angles at time point a and time point b, respectively. The preset historical time period can be the most recent preset time period, such as the last 24 hours. Therefore, the angle subsequence within the preset historical time period corresponding to a data point can be understood as the angle subsequence within the last 24 hours corresponding to the data point, and this angle subsequence includes the angle corresponding to the current data point. For example, if the data sampling period is 1 hour, the angle subsequence includes 24 angles, and the Gram angle difference field matrix is ​​a 24×24 matrix. Specifically, the Gram angle difference field matrices of the seven gases are used as the seven channels of the image, stacked to form a multi-channel temporal feature map.

[0046] In this embodiment, each data point is processed in the manner described above to obtain a multi-channel time series feature map corresponding to each data point, thereby obtaining a multi-channel time series feature map sequence.

[0047] S130, the multi-channel feature map sequence is input into the convolutional neural network to obtain the first global feature sequence and the first risk value sequence.

[0048] The convolutional neural network includes convolutional layers, a backbone network, global average pooling, fully connected layers, and a normalized exponential function. The backbone network is obtained by training a pre-trained EfficientNet-B0 network as follows: freezing the parameters of the first s layers, thawing the parameters of the remaining layers, and training the thawed layer parameters based on the oil chromatogram data sequence, where s is an integer greater than or equal to 1 and less than the total number of layers in the EfficientNet-B0 network. The convolutional layers can be 1×1 convolutional layers. The EfficientNet-B0 network includes an initial convolutional layer (e.g., Conv3×3), multiple feature extraction layers (e.g., MBConv), and a classification head layer. The pre-trained EfficientNet-B0 network can be a network trained on the original EfficientNet-B0 network based on an image classification dataset. Specifically, the pre-trained EfficientNet-B0 network can be trained by freezing the parameters of the first s layers (e.g., s=5), unfreezing the parameters of the remaining layers, setting an initial learning rate, using the Adam optimizer, and training the parameters of the unfrozen layers using oil chromatogram data sequences.

[0049] Optionally, the method for inputting the multi-channel feature map sequence into a convolutional neural network to obtain the first global feature sequence and the first risk value sequence can be as follows: Perform the following operations on each multi-channel feature map in the multi-channel feature map sequence to obtain the first global feature sequence and the first risk value sequence: Perform channel dimensionality reduction processing on the multi-channel feature map using a convolutional layer to obtain a three-channel feature map, and perform standardization processing on the three-channel feature map; input the standardized three-channel feature map into the backbone network to output a high-dimensional deep feature map; input the high-dimensional deep feature map into global average pooling to output the first global feature; input the first global feature into a fully connected layer to output the transformer operating state category; input the transformer operating state category into a normalized exponential function to output the probability of each operating state category; perform a weighted summation of the probabilities of each operating state category to obtain the first risk value.

[0050] In this embodiment, a 1×1 convolutional layer is used to perform stationary convolution on the multi-channel feature map to reduce the channel dimensionality, compressing the 7-channel feature map into a 3-channel feature map. This process only changes the number of channels, without changing the height and width of the feature map. The transformer operating state categories can include five types: normal state, low temperature overheating, medium temperature overheating, high temperature overheating, and discharge fault. The normalization exponential function can be the Softmax function.

[0051] Specifically, the formula for calculating the weighted sum of the probabilities of each operating state category can be expressed as: ,in, Indicates time The first risk value, Indicates the first The category risk weights for each running state, where s represents the running state category index. These correspond to normal state, low temperature overheating, medium temperature overheating, high temperature overheating, and discharge fault, respectively. Indicates the time of oil-immersed transformer The probability of being in the s-th running state. In this application scenario, The values ​​are determined based on the degree of impact of different operating conditions on the insulation system and the safety of the oil-immersed transformer itself. Under normal conditions, only background aging exists, resulting in the lowest risk. Low-temperature overheating indicates localized heating or mild thermal defects, but the immediate damage to the insulation is limited. Medium-temperature overheating indicates a deepening of thermal faults and a significant increase in the insulation aging rate. High-temperature overheating reflects higher hotspot temperatures, potentially accompanied by oil cracking and insulation paper deterioration, significantly increasing the risk. Discharge faults indicate the possible presence of partial discharge, spark discharge, and arc discharge within the equipment, posing a stronger destructive force to the insulation structure and the highest risk. Exemplary values ​​are taken as follows: 0.05 for normal conditions, 0.30 for low-temperature overheating, 0.55 for medium-temperature overheating, 0.80 for high-temperature overheating, and 0.90 for discharge faults, to reflect the monotonically increasing relationship of "normal condition – progressively worsening thermal faults – highest risk for discharge faults."

[0052] S140, determine the second risk value sequence based on the first risk value sequence and the operating environment data sequence.

[0053] In this embodiment, the first risk value sequence and the operating environment data sequence are time-axis aligned, meaning that the elements of the sequences correspond one-to-one. The method for determining the second risk value sequence based on the first risk value sequence and the operating environment data sequence can be: modifying the first risk value sequence based on the operating environment data sequence to obtain the second risk value sequence.

[0054] Optionally, the method for determining the second risk value sequence based on the first risk value sequence and the operating environment data sequence can be as follows: For each data point in the winding hotspot temperature sequence and the partial discharge amplitude sequence, obtain the winding hotspot temperature subsequence and the partial discharge amplitude subsequence within a preset historical period corresponding to the data point; determine the thermal risk correction factor based on the winding hotspot temperature subsequence, and determine the partial discharge risk correction factor based on the partial discharge amplitude subsequence; perform logarithmic compression on the thermal risk correction factor and the partial discharge risk correction factor respectively to obtain the thermal risk characterization quantity and the partial discharge risk characterization quantity; wherein, the thermal risk characterization quantity and the partial discharge risk characterization quantity of each data point constitute the thermal risk characterization quantity sequence and the partial discharge risk characterization quantity sequence respectively; and determine the second risk value sequence based on the first risk value sequence, the thermal risk characterization quantity sequence, and the partial discharge risk characterization quantity sequence.

[0055] In this sequence, the time axes of the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence are aligned, meaning there is a one-to-one correspondence between the elements of these sequences. The logarithmic compression process can be achieved by transforming the thermal risk correction factor and the partial discharge risk correction factor using a logarithmic function to obtain the thermal risk characterization and the partial discharge risk characterization, respectively. For example: y = klog(x), where y is the thermal risk characterization or the partial discharge risk characterization, x is the thermal risk correction factor or the partial discharge correction factor, and k is a preset coefficient.

[0056] The preset historical time period can be the most recent preset time period, such as the most recent 24 hours. Therefore, the winding hotspot temperature subsequence and partial discharge amplitude subsequence within the preset historical time period corresponding to the data point can be understood as: the winding hotspot temperature subsequence and partial discharge amplitude subsequence within the most recent 24 hours corresponding to the data point. The winding hotspot temperature subsequence includes the winding hotspot temperature corresponding to the current data point, and the partial discharge amplitude subsequence includes the partial discharge amplitude corresponding to the current data point. For example, if the data acquisition period is 1 hour, the winding hotspot temperature subsequence includes 24 winding hotspot temperatures, and the partial discharge amplitude subsequence includes 24 partial discharge amplitudes.

[0057] One method for determining the thermal risk correction factor based on the winding hotspot temperature subsequence is as follows: The winding hotspot temperature subsequence is processed using the Arrhenius aging mechanism to obtain the thermal risk correction factor, expressed as: ;in, Indicates time (i.e., data points) The thermal risk correction factor, Indicates time The winding hot spot temperature over the past m hours, where m represents the index of each data point in the winding hot spot temperature subsequence, and B represents the insulation paper thermal aging activation energy correlation constant, with a value of 15000. The reference temperature is 98°C in this invention.

[0058] The partial discharge risk correction factor, based on the partial discharge amplitude subsequence, can be calculated using the following formula: ;in, Indicates time Partial discharge risk correction factor Indicates time The amplitude of partial discharge over the past m hours, Indicates the partial discharge threshold. Indicates the weighting index. This indicates an indicator function (it takes a value of 1 when the condition is met, and 0 otherwise). In this application scenario, It is the dividing line between "normal allowable discharge" and "abnormal discharge". The present invention takes a value of 100pC. When the discharge exceeds 100pC, the discharge type changes from background discharge to developmental discharge, which has a cumulative effect on the deterioration of the insulation structure. Selecting 100pC as the threshold can suppress noise interference while maintaining a certain sensitivity to early defects, thereby achieving a balance between "not over-reporting false alarms and not under-reporting" in the operation risk assessment. To characterize the nonlinear amplification effect of partial discharge amplitude on risk, this invention uses a power function to model the discharge amplitude, thereby achieving the following: as the discharge amplitude increases, the discharge channel energy, local electric field distortion, and insulation damage rate all exhibit a nonlinear growth relationship; and the degree of nonlinearity is usually greater than the linear relationship but less than the exponential explosion relationship. Therefore, this invention exemplarily uses 1.5, which can reflect the amplification effect of high-amplitude discharge on risk, while taking into account both sensitivity and stability.

[0059] The formula for determining the second risk value sequence based on the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence can be expressed as: ,in, Indicates time The second risk value indicates that the higher the risk of operation of the oil-immersed transformer at that moment. This represents the Sigmoid function, where a, b, and c represent the weighting coefficients of the first risk value, the thermal risk characterization value, and the partial discharge risk characterization value, respectively. Indicates time Thermal risk characterization quantity, Indicates time The partial discharge risk characterization parameters are as follows. In this application scenario, the first risk value is extracted from oil chromatography data using a deep learning model. It directly reflects the combined state of thermal and discharge faults inside the oil-immersed transformer and is a characterization of internal equipment defects. Therefore, it should dominate the overall risk assessment. The thermal risk characterization parameter mainly describes the accelerating effect of temperature on the insulation aging rate and supplements the long-term degradation process. The partial discharge risk characterization parameter reflects the activity level of local electrical defects and corrects for sudden risks. Therefore, the contribution relationship of the three parameters should satisfy the following condition: For example, a=2, b=0.8, c=0.6, to ensure that changes in operational risks can be effectively distinguished.

[0060] S150, based on the first global feature sequence and the operating environment data sequence, a gated recurrent unit network is used to perform risk prediction to obtain at least one second predicted risk value.

[0061] Here, at least one second predicted risk value can be understood as at least one second predicted risk value at a future time, such as: the second predicted risk value for the next 1 hour, the second predicted risk value for the next 2 hours, ..., the second predicted risk value for the next P hours, etc. The Gated Recurrent Unit (GRU) network has the ability to remember historical temporal change patterns and can capture the evolution trend and temporal dependencies of the transformer state over time.

[0062] In this embodiment, the method of risk prediction based on the first global feature sequence and the operating environment data sequence using a gated recurrent unit network can be as follows: firstly, based on the first global feature sequence and the operating environment data sequence, a gated recurrent unit network is used to predict at least one future time's global features and operating environment data, and then based on the predicted global features and operating environment data, a second predicted risk value for the future time is determined.

[0063] Optionally, the method for obtaining at least one second predicted risk value by using a gated recurrent unit network to perform risk prediction based on the first global feature sequence and the operating environment data sequence can be as follows: concatenating each first global feature in the first global feature sequence with the corresponding operating environment data to obtain a fused feature sequence; inputting the fused feature sequence into the gated recurrent unit network for recursive prediction to obtain at least one predicted fused feature; wherein the predicted fused feature includes predicted global features and predicted operating environment data; wherein the predicted operating environment data includes predicted winding hotspot temperature and predicted partial discharge amplitude; inputting the predicted global features into the convolutional neural network to obtain a first predicted risk value; and determining at least one second predicted risk value based on the first predicted risk value, the predicted operating environment data, and the operating environment data sequence.

[0064] The process of concatenating each first global feature in the first global feature sequence with the corresponding operating environment data can be understood as concatenating each first global feature in the first global feature sequence with the winding hot spot temperature and partial discharge amplitude at the corresponding time to obtain a fused feature sequence.

[0065] In this embodiment, the fused feature sequence is input into a gated recurrent unit network to predict the fused feature at the next future time. The predicted fused feature is then incorporated into the fused feature sequence and input into the gated recurrent unit network to predict the fused feature at the next future time. This process is repeated to obtain the predicted fused features for P future time periods.

[0066] Specifically, the method for inputting the predicted global features into the convolutional neural network to obtain the first predicted risk value can be as follows: inputting the predicted global features into a fully connected layer to output the transformer operating state category; inputting the transformer operating state category into a normalized exponential function to output the probability of each operating state category; and weighting and summing the probabilities of each operating state category to obtain the first predicted risk value.

[0067] Specifically, the method for determining at least one second predicted risk value based on the first predicted risk value, predicted operating environment data, and operating environment data sequence can be as follows: determining the predicted thermal risk characterization quantity based on the winding hot spot temperature sequence and the predicted winding hot spot temperature; determining the predicted partial discharge risk characterization quantity based on the partial discharge amplitude sequence and the predicted partial discharge amplitude; and determining the second predicted risk value based on the first predicted risk value, the predicted thermal risk characterization quantity, and the predicted partial discharge risk characterization quantity.

[0068] The process of determining the predicted thermal risk characterization quantity based on the winding hot spot temperature sequence and the predicted winding hot spot temperature can be as follows: merging the predicted winding hot spot temperature into the winding hot spot temperature sequence to obtain a new winding hot spot temperature sequence; obtaining the winding hot spot temperature subsequence within the most recent historical period corresponding to the predicted winding hot spot temperature from the new winding hot spot temperature sequence; processing the winding hot spot temperature subsequence using the Arrhenius aging mechanism to obtain the predicted thermal risk correction factor; and performing logarithmic compression on the predicted thermal risk correction factor to obtain the predicted thermal risk characterization quantity.

[0069] The process of determining the predicted partial discharge risk characterization quantity based on the partial discharge amplitude sequence and the predicted partial discharge amplitude can be as follows: merging the predicted partial discharge amplitude into the partial discharge amplitude sequence to obtain a new partial discharge amplitude sequence; obtaining the partial discharge amplitude subsequence within the most recent historical period corresponding to the predicted partial discharge amplitude from the new partial discharge amplitude sequence, processing the partial discharge amplitude subsequence in the manner described in the above embodiment to obtain the predicted partial discharge risk correction factor; and performing logarithmic compression on the predicted partial discharge risk correction factor to obtain the predicted partial discharge risk characterization quantity.

[0070] The formula for determining the second predicted risk value based on the first predicted risk value, the predicted thermal risk characterization value, and the predicted partial discharge risk characterization value can be expressed as follows: ,in, This represents the first predicted risk value. This represents the second predicted risk value. This represents the Sigmoid function, where a, b, and c represent the weighting coefficients of the first predicted risk value, the predicted thermal risk characterization value, and the predicted partial discharge risk characterization value, respectively. This represents a quantity that indicates the predicted thermal risk. This represents a predictive measure of partial discharge risk.

[0071] S160, Perform curve fitting on the second risk value sequence and at least one second predicted risk value to obtain a risk curve.

[0072] Among them, the risk curve is used to characterize the risk change trend of oil-immersed transformers.

[0073] Optional features include: displaying a risk curve.

[0074] The curve segments corresponding to the second risk value sequence and at least one second predicted risk value in the risk curve are displayed in different styles.

[0075] The style may include at least one of the following: color, thickness, and solid / dashed lines, for example, the curve segment corresponding to the second risk value sequence is displayed with a solid line, and the curve segment corresponding to at least one second predicted risk value is displayed with a dashed line.

[0076] In this embodiment, curve fitting of the second risk value sequence and at least one second predicted risk value can be understood as: incorporating at least one second predicted risk value into the second risk value sequence to obtain a new second risk value sequence, and then performing curve fitting on the new second risk value sequence.

[0077] Optionally, the future risk evolution trend can be indicated by arrow direction and color change in the risk curve, allowing users to see the risk evolution trend more intuitively.

[0078] Optionally, the first risk value, thermal risk characterization, and partial discharge risk characterization used to generate the second risk value can be displayed separately; as well as the first predicted risk value, predicted thermal risk characterization, and predicted partial discharge risk characterization used to generate the second predicted risk value. The contribution ratio of these three components is displayed graphically.

[0079] The technical solution of this embodiment involves acquiring oil chromatography data sequences and operating environment data sequences, and preprocessing these sequences; constructing a Gram angle difference field matrix based on the preprocessed oil chromatography data sequences, and stacking the Gram angle difference field matrices to generate a multi-channel feature map sequence; inputting the multi-channel feature map sequence into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence; determining a second risk value sequence based on the first risk value sequence and the operating environment data sequence; performing risk prediction using a gated recurrent unit network based on the first global feature sequence and the operating environment data sequence to obtain at least one second predicted risk value; and performing curve fitting on the second risk value sequence and at least one second predicted risk value to obtain a risk curve; wherein, the risk curve is used to characterize the risk change trend of the oil-immersed transformer. The operational risk prediction method for oil-immersed transformers provided in this invention constructs a Gram angle difference field matrix to transform oil chromatography data into a feature map suitable for convolutional neural network processing, thus preserving temporal dependencies; it determines a second risk value sequence by using operating environment data and a first risk value sequence, thereby improving the accuracy of risk value determination; and it performs risk prediction based on a first global feature sequence and operating environment data to predict risk values ​​at multiple future times, thereby improving prediction accuracy.

[0080] Example 2

[0081] Figure 2 This is a schematic diagram of a device for predicting the operational risk of an oil-immersed transformer, provided as an embodiment of the present invention. Figure 2 As shown, the device specifically includes:

[0082] The data acquisition module 210 is used to acquire oil chromatography data sequences and operating environment data sequences, and to preprocess the oil chromatography data sequences and the operating environment data sequences.

[0083] The multi-channel feature map sequence generation module 220 is used to construct a Gram angle difference field matrix based on the preprocessed oil chromatography data sequence, and to stack the Gram angle difference field matrix to generate a multi-channel feature map sequence.

[0084] The first global feature sequence and first risk value sequence acquisition module 230 is used to input the multi-channel feature map sequence into a convolutional neural network to obtain the first global feature sequence and the first risk value sequence.

[0085] The second risk value sequence determination module 240 is used to determine a second risk value sequence based on the first risk value sequence and the operating environment data sequence.

[0086] The second predicted risk value determination module 250 is used to perform risk prediction based on the first global feature sequence and the operating environment data sequence using a gated recurrent unit network to obtain at least one second predicted risk value.

[0087] The risk curve acquisition module 260 is used to perform curve fitting on the second risk value sequence and the at least one second predicted risk value to obtain a risk curve; wherein, the risk curve is used to characterize the risk change trend of the oil-immersed transformer.

[0088] Optionally, the oil chromatographic data sequence includes the following gas concentration sequences dissolved in the oil: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide gas concentrations.

[0089] Optionally, the multi-channel feature map sequence generation module 220 is also used for:

[0090] The preprocessed gas concentration sequence is mapped to polar coordinate space to obtain the angle sequence;

[0091] For each data point in the angle sequence, obtain the angle subsequence within a preset historical time period corresponding to the data point;

[0092] Calculate the sine value of the pairwise angle difference in the angle subsequence to obtain the Gram angle difference field matrix corresponding to each gas.

[0093] The Gram difference field matrices corresponding to each gas are stacked to obtain the multi-channel feature map corresponding to the data point; wherein, the multi-channel feature map corresponding to each data point constitutes a multi-channel feature map sequence.

[0094] Optionally, the convolutional neural network includes convolutional layers, a backbone network, global average pooling, fully connected layers, and a normalized exponential function; wherein, the backbone network is obtained by training a pre-trained EfficientNet-B0 network in the following manner: freezing the parameters of the first s layers, thawing the parameters of the remaining layers, and training the parameters of the thawed layers based on the oil chromatogram data sequence, where s is an integer greater than or equal to 1 and less than the total number of layers in the EfficientNet-B0 network.

[0095] Optionally, the first global feature sequence and first risk value sequence acquisition module 230 is also used for:

[0096] The following operations are performed on each multi-channel feature map in the multi-channel feature map sequence to obtain the first global feature sequence and the first risk value sequence:

[0097] The convolutional layer is used to perform channel dimensionality reduction on the multi-channel feature map to obtain a three-channel feature map, and the three-channel feature map is then standardized.

[0098] The standardized three-channel feature map is input into the backbone network, and a high-dimensional deep feature map is output.

[0099] The high-dimensional deep feature map is input into the global average pooling, and the first global feature is output.

[0100] The first global feature is input into the fully connected layer, and the transformer operating state category is output.

[0101] Input the transformer operating state category into the normalized exponential function, and output the probability of each operating state category;

[0102] The probabilities of each of the aforementioned operating state categories are weighted and summed to obtain the first risk value.

[0103] Optionally, the operating environment data sequence includes a winding hotspot temperature sequence and a partial discharge amplitude sequence.

[0104] Optionally, the second risk value sequence determination module 240 is also used for:

[0105] For each data point in the winding hotspot temperature sequence and the partial discharge amplitude sequence, obtain the winding hotspot temperature sub-sequence and the partial discharge amplitude sub-sequence within a preset historical time period corresponding to the data point;

[0106] A thermal risk correction factor is determined based on the winding hot spot temperature subsequence, and a partial discharge risk correction factor is determined based on the partial discharge amplitude subsequence.

[0107] Logarithmic compression is performed on the thermal risk correction factor and the partial discharge risk correction factor to obtain thermal risk characterization and partial discharge risk characterization, respectively; wherein, the thermal risk characterization and partial discharge risk characterization of each data point constitute the thermal risk characterization sequence and the partial discharge risk characterization sequence, respectively.

[0108] A second risk value sequence is determined based on the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence; wherein the time axes of the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence are aligned.

[0109] Optionally, the second predicted risk value determination module 250 is also used for:

[0110] Each first global feature in the first global feature sequence is concatenated with its corresponding runtime environment data to obtain a fused feature sequence;

[0111] The fused feature sequence is input into a gated recurrent unit network for recursive prediction to obtain at least one predicted fused feature; wherein, the predicted fused feature includes predicted global features and predicted operating environment data; wherein, the predicted operating environment data includes predicted winding hot spot temperature and predicted partial discharge amplitude;

[0112] The predicted global features are input into the convolutional neural network to obtain a first predicted risk value;

[0113] The at least one second predicted risk value is determined based on the first predicted risk value, the predicted operating environment data, and the operating environment data sequence.

[0114] Optionally, the second predicted risk value determination module 250 is also used for:

[0115] The predicted thermal risk characterization quantity is determined based on the winding hot spot temperature sequence and the predicted winding hot spot temperature.

[0116] Based on the partial discharge amplitude sequence and the predicted partial discharge amplitude, a predicted partial discharge risk characterization quantity is determined.

[0117] A second predicted risk value is determined based on the first predicted risk value, the predicted thermal risk characterization, and the predicted partial discharge risk characterization.

[0118] Optionally, it also includes: a display module, used for:

[0119] The risk curve is displayed; wherein the curve segment corresponding to the second risk value sequence and the curve segment corresponding to the at least one second predicted risk value are displayed in different styles.

[0120] The above-described products can perform the methods provided in any embodiment of the present invention and have the corresponding functional modules for performing the methods.

[0121] Example 3

[0122] Figure 3 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0123] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory 12 or a random access memory 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 12 or loaded from storage unit 18 into the random access memory 13. The random access memory 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, read-only memory 12, and random access memory 13 are interconnected via a bus 14. An input / output interface 15 is also connected to the bus 14.

[0124] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0125] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing units, graphics processing units, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, digital signal processors, and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the operational risk prediction method for oil-immersed transformers.

[0126] In some embodiments, the method for predicting the operational risks of an oil-immersed transformer can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via read-only memory 12 and / or communication unit 19. When the computer program is loaded into random access memory 13 and executed by processor 11, one or more steps of the method for predicting the operational risks of an oil-immersed transformer described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the method for predicting the operational risks of an oil-immersed transformer by any other suitable means (e.g., by means of firmware).

[0127] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays, application-specific integrated circuits (ASICs), application-specific standard products (ASICs), systems-on-a-chip (SoCs), payload programmable logic devices, computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0129] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0130] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube or liquid crystal display) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0131] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0132] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to address the shortcomings of traditional physical hosts and virtual private servers, such as high management difficulty and weak business scalability.

[0133] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0134] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for predicting operational risks of oil-immersed transformers according to any embodiment of the invention.

[0135] In the implementation of a computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for predicting operational risks of oil-immersed transformers, characterized in that, include: Obtain oil chromatography data sequences and operating environment data sequences, and preprocess the oil chromatography data sequences and the operating environment data sequences; A Gram angle difference field matrix is ​​constructed based on the preprocessed oil chromatography data sequence, and the Gram angle difference field matrix is ​​stacked to generate a multi-channel feature map sequence. The multi-channel feature map sequence is input into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence. A second risk value sequence is determined based on the first risk value sequence and the operating environment data sequence; Based on the first global feature sequence and the operating environment data sequence, a gated recurrent unit network is used to perform risk prediction to obtain at least one second predicted risk value. A risk curve is obtained by curve fitting the second risk value sequence and the at least one second predicted risk value; wherein the risk curve is used to characterize the risk change trend of the oil-immersed transformer.

2. The method according to claim 1, characterized in that, The oil chromatographic data sequence includes the following gas concentration sequences dissolved in the oil: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide gas concentrations; The process involves constructing a Gram angle difference field matrix based on the preprocessed oil chromatography data sequence, and stacking the Gram angle difference field matrix to generate a multi-channel feature map sequence, including: The preprocessed gas concentration sequence is mapped to polar coordinate space to obtain the angle sequence; For each data point in the angle sequence, obtain the angle subsequence within a preset historical time period corresponding to the data point; Calculate the sine value of the pairwise angle difference in the angle subsequence to obtain the Gram angle difference field matrix corresponding to each gas. The Gram difference field matrices corresponding to each gas are stacked to obtain the multi-channel feature map corresponding to the data point; wherein, the multi-channel feature map corresponding to each data point constitutes a multi-channel feature map sequence.

3. The method according to claim 1, characterized in that, The convolutional neural network includes convolutional layers, a backbone network, global average pooling, fully connected layers, and a normalized exponential function; wherein, the backbone network is obtained by training a pre-trained EfficientNet-B0 network in the following manner: freezing the parameters of the first s layers, thawing the parameters of the remaining layers, and training the parameters of the thawed layers based on the oil chromatogram data sequence, where s is an integer greater than or equal to 1 and less than the total number of layers in the EfficientNet-B0 network; The step of inputting the multi-channel feature map sequence into a convolutional neural network to obtain a first global feature sequence and a first risk value sequence includes: The following operations are performed on each multi-channel feature map in the multi-channel feature map sequence to obtain the first global feature sequence and the first risk value sequence: The convolutional layer is used to perform channel dimensionality reduction on the multi-channel feature map to obtain a three-channel feature map, and the three-channel feature map is then standardized. The standardized three-channel feature map is input into the backbone network, and a high-dimensional deep feature map is output. The high-dimensional deep feature map is input into the global average pooling, and the first global feature is output. The first global feature is input into the fully connected layer, and the transformer operating state category is output. Input the transformer operating state category into the normalized exponential function, and output the probability of each operating state category; The probabilities of each of the aforementioned operating state categories are weighted and summed to obtain the first risk value.

4. The method according to claim 1, characterized in that, The operating environment data sequence includes the winding hot spot temperature sequence and the partial discharge amplitude sequence; The step of determining the second risk value sequence based on the first risk value sequence and the operating environment data sequence includes: For each data point in the winding hotspot temperature sequence and the partial discharge amplitude sequence, obtain the winding hotspot temperature sub-sequence and the partial discharge amplitude sub-sequence within a preset historical time period corresponding to the data point; A thermal risk correction factor is determined based on the winding hot spot temperature subsequence, and a partial discharge risk correction factor is determined based on the partial discharge amplitude subsequence. Logarithmic compression is performed on the thermal risk correction factor and the partial discharge risk correction factor to obtain thermal risk characterization and partial discharge risk characterization, respectively; wherein, the thermal risk characterization and partial discharge risk characterization of each data point constitute the thermal risk characterization sequence and the partial discharge risk characterization sequence, respectively. A second risk value sequence is determined based on the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence; wherein the time axes of the first risk value sequence, the thermal risk characterization sequence, and the partial discharge risk characterization sequence are aligned.

5. The method according to claim 4, characterized in that, Based on the first global feature sequence and the operating environment data sequence, a gated recurrent unit network is used for risk prediction to obtain at least one second predicted risk value, including: Each first global feature in the first global feature sequence is concatenated with its corresponding runtime environment data to obtain a fused feature sequence; The fused feature sequence is input into the gated cyclic unit network for recursive prediction to obtain at least one predicted fused feature; wherein, the predicted fused feature includes predicted global features and predicted operating environment data; wherein, the predicted operating environment data includes predicted winding hot spot temperature and predicted partial discharge amplitude; The predicted global features are input into the convolutional neural network to obtain a first predicted risk value; The at least one second predicted risk value is determined based on the first predicted risk value, the predicted operating environment data, and the operating environment data sequence.

6. The method according to claim 5, characterized in that, Determining the at least one second predicted risk value based on the first predicted risk value, the predicted operating environment data, and the operating environment data sequence includes: The predicted thermal risk characterization quantity is determined based on the winding hot spot temperature sequence and the predicted winding hot spot temperature. Based on the partial discharge amplitude sequence and the predicted partial discharge amplitude, a predicted partial discharge risk characterization quantity is determined. A second predicted risk value is determined based on the first predicted risk value, the predicted thermal risk characterization, and the predicted partial discharge risk characterization.

7. The method according to claim 1, characterized in that, Also includes: The risk curve is displayed; wherein the curve segment corresponding to the second risk value sequence and the curve segment corresponding to the at least one second predicted risk value are displayed in different styles.

8. A device for predicting operational risks of oil-immersed transformers, characterized in that, include: The data acquisition module is used to acquire oil chromatography data sequences and operating environment data sequences, and to preprocess the oil chromatography data sequences and the operating environment data sequences. The multi-channel feature map sequence generation module is used to construct a Gram angle difference field matrix based on the preprocessed oil chromatography data sequence, and to stack the Gram angle difference field matrix to generate a multi-channel feature map sequence. The module for obtaining the first global feature sequence and the first risk value sequence is used to input the multi-channel feature map sequence into a convolutional neural network to obtain the first global feature sequence and the first risk value sequence. The second risk value sequence determination module is used to determine a second risk value sequence based on the first risk value sequence and the operating environment data sequence. The second predicted risk value determination module is used to perform risk prediction based on the first global feature sequence and the operating environment data sequence using a gated recurrent unit network to obtain at least one second predicted risk value. The risk curve acquisition module is used to perform curve fitting on the second risk value sequence and the at least one second predicted risk value to obtain a risk curve; wherein the risk curve is used to characterize the risk change trend of the oil-immersed transformer.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the operational risk prediction method for oil-immersed transformers according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method for predicting the operational risk of an oil-immersed transformer as described in any one of claims 1-7.

Images