Transformer internal insulation state abnormity detection method, device, equipment and medium

Abstract

The invention relates to the technical field of transformer anomaly detection, and discloses a transformer internal insulation state anomaly detection method, device and equipment and a medium, and the method comprises the steps: calculating a correlation factor set between data sequences collected by measurement points at any two different positions in a transformer; establishing an input vector; obtaining a predicted value of the data collected by each measuring point in the next day through an informer model; constructing a prediction sequence based on the prediction values of the continuous preset days and the association factor set, and constructing a real data sequence based on the real data actually collected by each measurement point of the continuous preset days and the association factor set; the difference value between the prediction sequence and the real data sequence is calculated, whether the internal insulation state of the transformer is abnormal or not is judged based on the difference value, and the accuracy of internal insulation state detection of the transformer is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer abnormal detection, and particularly to a method, device, equipment and medium for detecting abnormal internal insulation state of a transformer. Background Art

[0002] The oil temperature of a transformer can reflect the operating state of the transformer and plays a key role in monitoring the health status of the transformer and judging the internal insulation state of the transformer. However, the transformer oil temperature sensor is usually in a complex environment and has been in a high-temperature and strong-magnetic working environment for a long time, resulting in a reduction in its service life and operating reliability, and numerical deviation may occur, which has a negative impact on the reliability of the sensor sensing data. In addition, it is of great significance to detect the abnormal internal insulation state of the transformer in a timely manner based on the data of the oil temperature sensor for monitoring the operating state of the transformer.

[0003] For the detection of abnormal sensor data, currently, abnormal detection can be performed through the statistical characteristics of the data, and the variance characteristics of the data segment are calculated by using the Tukey method for judgment; there are also some abnormal detection methods that use some clustering algorithms (such as Gaussian mixture model and DBSCAN) to convert the task into data sample density estimation. If the data is located in a low-density area, it is judged as data anomaly. The above methods can all achieve the abnormal detection of data, but they cannot perform real-time detection of the data. The related technology introduces the method of deep learning and proposes that the method based on LSTM can achieve the discrimination of oil temperature data. However, the LSTM model is not good at processing long segment data and the detection accuracy is low. Summary of the Invention

[0004] In view of this, the present invention provides a method, device, equipment and medium for detecting abnormal internal insulation state of a transformer to solve the problem of the detection accuracy of the internal insulation state of the transformer in the prior art.

[0005] In a first aspect, the present invention provides a method for detecting abnormal internal insulation state of a transformer, including: calculating an association factor set between data sequences collected at measurement points at any two different positions inside the transformer; constructing an input vector according to the real data collected at each measurement point on the current date, the real data collected in a preset number of days before the current date, and the association factor set; inputting the input vector into an informer model, and obtaining the predicted values of the data collected at each measurement point on the next day through the informer model; constructing a prediction sequence based on the predicted values of consecutive preset days and the association factor set, and constructing a real data sequence based on the real data actually collected at each measurement point for consecutive preset days and the association factor set; calculating the difference value between the prediction sequence and the real data sequence, and judging whether the internal insulation state of the transformer is abnormal based on the difference value.

[0006] Optionally, the set of correlation factors between the data sequences collected at any two different positions inside the computing transformer includes: calculating the absolute value, increment value, and growth rate value between the data sequences collected at any two different positions inside the computing transformer; determining the set of correlation factors based on the obtained absolute value, increment value, and growth rate value.

[0007] Optionally, the formula for calculating the absolute value between the data sequences collected at any two different positions inside the transformer is:

[0008]

[0009] In the formula, L 1 represents the absolute value, x it and x jt are the data sequences of measurement point i and measurement point j at time t respectively, and T is the data acquisition time window.

[0010] Optionally, the formula for calculating the increment value between the data sequences collected at any two different positions inside the transformer is:

[0011]

[0012] In the formula, L 2 represents the increment value, y it and y jt are the change amounts of the data sequences of measurement point i and measurement point j at time t relative to time t - 1 respectively.

[0013] Optionally, the formula for calculating the growth rate value between the data sequences collected at any two different positions inside the transformer is:

[0014]

[0015] In the formula, L 3 represents the growth rate value, z it and z jt are the change amplitudes of the increment values of the data sequences of measurement point i and measurement point j at time t relative to time t - 1 respectively.

[0016] Optionally, determining the set of correlation factors based on the obtained absolute value, increment value, and growth rate value includes:

[0017] Calculating the correlation factors of any two measurement points according to the following calculation formula:

[0018]

[0019] In the formula, w 1 、w 2 and w 3is the weight coefficient, and H is the correlation factor between any two measurement points;

[0020] Construct a correlation factor set according to the correlation factors between any two measurement points, and the construction formula is:

[0021] x L ={H 1 , H 2 ,,, H m}

[0022] In the formula, x L is the correlation factor set, and H m is the m-th correlation factor.

[0023] Optionally, before constructing the input vector according to the real data collected at each measurement point on the current date, the real data collected in the preset number of days before the current date, and the correlation factor set, it includes: performing data cleaning on the collected real data; performing normalization processing on the cleaned real data.

[0024] Optionally, calculating the difference value between the prediction sequence and the real data sequence includes: calculating the difference value between the prediction sequence and the real data sequence using the L2 norm.

[0025] In a second aspect, the present invention provides a transformer internal insulation state abnormal detection device, including: a factor calculation module for calculating a correlation factor set between data sequences collected at any two different positions inside the transformer; a vector construction module for constructing an input vector according to the real data collected at each measurement point on the current date, the real data collected in the preset number of days before the current date, and the correlation factor set; a prediction module for inputting the input vector into the informer model to obtain predicted values of the data collected at each measurement point on the next day through the informer model; a sequence construction module for constructing a prediction sequence based on the predicted values of consecutive preset days and the correlation factor set, and constructing a real data sequence based on the real data actually collected at each measurement point for consecutive preset days and the correlation factor set; an abnormality determination module for calculating the difference value between the prediction sequence and the real data sequence, and determining whether the internal insulation state of the transformer is abnormal based on the difference value.

[0026] In a third aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the transformer internal insulation state abnormal detection method according to the first aspect or any corresponding embodiment thereof.

[0027] Fourthly, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to make a computer execute the method for detecting abnormal internal insulation state of a transformer according to the first aspect or any corresponding embodiment thereof described above.

[0028] As can be seen from the above technical solutions, the embodiments of the present invention have the following advantages:

[0029] A method, device, equipment and medium for detecting abnormal internal insulation state of a transformer provided by the present invention calculate a set of correlation factors between data sequences collected at measurement points at any two different positions inside the transformer; construct an input vector according to the real data collected at each measurement point on the current date, the real data collected on the preset number of days before the current date, and the set of correlation factors; input the input vector into the informer model, and obtain the predicted values of the data collected at each measurement point on the next day through the informer model; construct a prediction sequence based on the predicted values of the continuous preset number of days and the set of correlation factors, and construct a real data sequence based on the real data actually collected at each measurement point for the continuous preset number of days and the set of correlation factors; calculate the difference value between the prediction sequence and the real data sequence, and judge whether the internal insulation state of the transformer is abnormal based on the difference value. Using the informer model for prediction can reduce the computational complexity, significantly improve the information extraction ability of long sequences, represent the correlation of data at different measurement points through the set of correlation factors, and input it as part of the input vector to comprehensively extract the same data trend from the data at different measurement points, so as to make certain corrections to the input data, increase the accuracy of the predicted data, greatly improve the performance of long-term sequence prediction, improve the accuracy of the predicted data, and further increase the accuracy of abnormal data detection. Description of the Drawings

[0030] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0031] Figure 1 It is a schematic flow chart of a method for detecting abnormal internal insulation state of a transformer according to an embodiment of the present invention;

[0032] Figure 2 It is a schematic flow chart of another method for detecting abnormal internal insulation state of a transformer according to an embodiment of the present invention;

[0033] Figure 3It is a data schematic diagram of abnormal continuous repeated values of the oil temperature sensor data in the embodiment of the present invention;

[0034] Figure 4 It is a data schematic diagram of abnormal fluctuations in the oil temperature sensor data in the embodiment of the present invention;

[0035] Figure 5 It is a data schematic diagram of abnormal outliers in the oil temperature sensor data in the embodiment of the present invention;

[0036] Figure 6 It is a data schematic diagram of abnormal drift in the oil temperature sensor data in the embodiment of the present invention;

[0037] Figure 7 It is a data schematic diagram of normal data of the oil temperature sensor in the embodiment of the present invention;

[0038] Figure 8 It is a structural block diagram of a transformer internal insulation state abnormal detection device in the embodiment of the present invention;

[0039] Figure 9 It is a hardware structure schematic diagram of a computer device in the embodiment of the present invention. Specific embodiments

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] Online monitoring data of transformer oil temperature plays an important role in judging the operation reliability of the transformer and monitoring the internal insulation state of the transformer. Due to the relatively complex environment where the oil temperature sensor is located, it is very easy to have a failure, which may cause the monitoring data to fail due to sensor failure, and the resulting abnormal data affects the judgment of the transformer state. Therefore, the present invention analyzes the abnormal types of sensor data and proposes a method for detecting abnormal internal insulation state of a transformer, which can identify abnormal states in long-term oil temperature monitoring data.

[0042] The embodiment of the present invention provides a method for detecting abnormal internal insulation state of a transformer, as shown in Figure 1 and Figure 2 shown, the method includes:

[0043] Step S101, calculate the correlation factor set between the data sequences collected at the measurement points at any two different positions inside the transformer.

[0044] Specifically, corresponding data sequences are collected by installing oil temperature sensors at measurement points in different positions. The oil temperature sensors are used to collect the top oil temperature at various positions of the transformer. Due to the spatial position differences of the measurement points, the collected oil temperature data varies due to the influence of spatio-temporal relationships, but the oil temperature of the same transformer has the same trend change and data fluctuation for these two reasons.

[0045] The associated factor set includes several associated factors, and the comprehensive distance between two measurement points is represented by the associated factors, that is, the correlation between the data measured at the two measurement points.

[0046] Step S102: Construct an input vector based on the real data collected at each measurement point on the current date, the real data collected in a preset number of days before the current date, and the associated factor set.

[0047] Exemplarily, if the preset number of days is 40 days, the input vector includes the real data collected at each measurement point on the current day, the real data collected in the preset number of days before the current date, and the associated factor set. The real data collected every day is partially extracted from the data segment in the monitoring data of the oil temperature sensor through a time sliding window.

[0048] The data collected at each measurement point every day consists of three parts, namely: raw data + location information + associated factor.

[0049] Step S103: Input the input vector into the Informer model, and obtain the predicted values of the data collected at each measurement point on the next day through the Informer model.

[0050] Specifically, the Informer model includes an encoder and a decoder. The encoder receives a long sequence input and uses the ProbSparse self-attention module and the self-attention distillation module to obtain a feature representation. The decoder receives a long sequence input and interacts with the encoded features through multi-head attention. First, for the Informer model, since the input vector lacks a method to represent the order of the input data sequence, in order to represent the order of the data, a position encoding vector needs to be added to the input vector, and the addition process is as follows:

[0051]

[0052]

[0053] pos represents the position of the current word in the sentence, d model represents the dimension of the encoding, and i ∈ [0, d model .

[0054] For the self-attention module of the Informer model, the starting point is that due to the sparsity of the self-attention probability distribution, only a few dot products contribute most of the attention. Therefore, only the top-ranked M(qi, K) (the sparsity measure of the i-th query) needs to be selected. First, randomly sample the key values, and then calculate the dot product of each query with the sampled keys. Then, calculate the difference between the attention of the dot product corresponding to each query and the uniform distribution, and then select the top u queries with the largest difference as the calculation of the attention probability, and the remaining values are filled with the average value.

[0055] Through the Informer model, the predicted values of the data collected at each measurement point on the next day can be predicted, and the correlation of the data at each measurement point can be represented by a pre-constructed set of correlation factors, which can make the predicted data more accurate.

[0056] Step S104, construct a prediction sequence based on the predicted values for consecutive preset days and the set of correlation factors, and construct a true data sequence based on the actual true data collected at each measurement point for consecutive preset days and the set of correlation factors.

[0057] For each measurement point, construct a prediction sequence using the predicted values obtained by the Informer model, and construct a true data sequence using the actually collected data. Constructing the prediction sequence and the true data sequence provides a comparison benchmark for subsequent anomaly detection and helps to discover anomalies.

[0058] Step S105, calculate the difference value between the prediction sequence and the true data sequence, and judge whether the internal insulation state of the transformer is abnormal based on the difference value.

[0059] Calculate the difference value between the prediction sequence and the true data sequence, and judge whether the data of each measurement point is abnormal by setting a threshold or using a statistical method. For each measurement point, calculate the difference value between the prediction sequence and the true data sequence, and L2 norm, mean square error (MSE) or other appropriate metrics can be used. Through the calculation of the difference value, it can be judged whether the data of each measurement point deviates from the expectation, so as to identify possible abnormal situations.

[0060] A method for detecting abnormal internal insulation state of a transformer provided by the present invention calculates a set of correlation factors between data sequences collected at measurement points at any two different positions inside the transformer; constructs an input vector according to the real data collected at each measurement point on the current date, the real data collected on a preset number of days before the current date, and the set of correlation factors; inputs the input vector into the Informer model, and obtains predicted values of the data collected at each measurement point on the next day through the Informer model; constructs a prediction sequence based on the predicted values of consecutive preset days and the set of correlation factors, and constructs a real data sequence based on the real data actually collected at each measurement point on consecutive preset days and the set of correlation factors; calculates the difference value between the prediction sequence and the real data sequence, and judges whether the internal insulation state of the transformer is abnormal based on the difference value. Using the Informer model for prediction can reduce the computational complexity, significantly improve the information extraction ability of long sequences, represent the correlation of data at different measurement points through the set of correlation factors, and input it as part of the input vector to synthesize the data of different measurement points to extract the same data trend, thereby making certain corrections to the input data, increasing the accuracy of the predicted data, greatly improving the performance of long-term sequence prediction, improving the accuracy of the predicted data, and further increasing the accuracy of abnormal data detection.

[0061] A method for detecting abnormal internal insulation state of a transformer according to an embodiment of the present invention combines the data correlation, historical information, and prediction ability of a machine learning model in its overall process, making the monitoring of abnormal data at the internal measurement points of the transformer more comprehensive and accurate.

[0062] In an optional embodiment, before step S102 of constructing an input vector according to the real data collected at each measurement point on the current date, the real data collected on a preset number of days before the current date, and the set of correlation factors, it includes:

[0063] Perform data cleaning on the collected real data;

[0064] Perform normalization processing on the cleaned real data.

[0065] Specifically, based on the data anomaly types of the oil temperature sensor, the data is preprocessed, and the preprocessing process includes data cleaning and normalization processing.

[0066] Through the analysis of the oil temperature sensor data samples and the summary of relevant research in the embodiments of the present invention, the data anomalies of the oil temperature sensor are divided into four types: continuous repeated value anomaly, fluctuation anomaly, outlier anomaly, and drift anomaly. As Figure 3As shown, the continuous repeated values of the oil temperature sensor data are abnormal. The oil temperature data should be a set of continuously fluctuating data. However, the sensor data shows that the oil temperature has remained at the same temperature for a long time, with no change or very little change. This may be due to a malfunction in the measurement part of the sensor, resulting in the sensor being unable to measure new oil temperature data. Its data upload module continues to upload the previous measurement value, causing the oil temperature data to remain in a non-fluctuating state for a long time. As Figure 4 As shown, the fluctuation of the oil temperature sensor data is abnormal. After a certain measurement time node, the volatility of the oil temperature sensor data changes suddenly. It can be seen from the figure that the data standard deviation becomes significantly larger. As Figure 5 As shown, the outliers of the oil temperature sensor data are abnormal. Values that are far from the average at other times suddenly appear in the oil temperature data. The reason for this situation may be a sensor failure or the actual state of the equipment. A huge change has indeed occurred, and a comprehensive discriminant analysis needs to be carried out based on the data at other times. As Figure 6 As shown, the data drift of the oil temperature sensor is abnormal. It can be seen that the average value of the data has a significant change, and the overall data has an obvious drift. Figure 7 This is the normal value of the oil temperature sensor data.

[0067] Regarding the types and characteristics of abnormal data, first collect the real data on-site to form the training set of the informer model composed of normal samples. It is necessary to clean the real data. If a certain data point will affect the subsequent algorithm process, the real data of this point should be estimated and replaced. Common methods for estimating the true value include interpolation method and replacement with adjacent valid values, etc. These methods can achieve good results in the case of non-large amounts of continuous abnormal data. When the data set is densely sampled, individual abnormal data can also be discarded with little impact on the original data information.

[0068] After that, perform z-score normalization on the data. Z-score normalization assumes that the data follows a normal distribution. First, estimate the population mean μ through the sample mean, estimate the population variance σ with the sample variance, and divide the data minus the mean by the variance to obtain the normalization result, as shown in the following formula:

[0069]

[0070] x is the data before normalization, and x’ is the data after normalization. This normalization method based on probability distribution can effectively increase the distance between data and further amplify the fluctuation of the data.

[0071] Through the analysis of the oil temperature data in the embodiments of the present invention, the characteristic differences between the normal oil temperature sensor data and various abnormal oil temperature sensor data in the time dimension are effectively mined. Based on this, the collected real-time data is regularized, avoiding directly inputting the collected abnormal data into the informer model, and improving the accuracy of the model.

[0072] In an optional embodiment, in step S101, calculating a set of correlation factors between data sequences collected at any two different positions inside the transformer, including:

[0073] Step S1011, calculating the absolute value, increment value, and growth rate value between the data sequences collected at any two different positions inside the transformer.

[0074] Specifically, by calculating the three Euclidean distances of the absolute value, increment value, and growth rate value between the data sequences collected at any two different positions inside the transformer, an index of similarity between the measurement points, that is, a correlation factor, is represented.

[0075] The calculation formula for the absolute value between the data sequences collected at any two different positions inside the transformer is:

[0076]

[0077] In the formula, L 1 represents the absolute value, x it and x jt are the data sequences of measurement point i and measurement point j at time t respectively, and T is the data acquisition time window.

[0078] The calculation formula for the increment value between the data sequences collected at any two different positions inside the transformer is:

[0079]

[0080] In the formula, L 2 represents the increment value, y it and y jt are the change amounts of the data sequences of measurement point i and measurement point j at time t relative to time t - 1 respectively.

[0081] The calculation formula for the growth rate value between the data sequences collected at any two different positions inside the transformer is:

[0082]

[0083] In the formula, L 3 represents the growth rate value, z it and z jt are the change amplitudes of the increment values of the data sequences of measurement point i and measurement point j at time t relative to time t - 1 respectively.

[0084] Step S1012, determining the set of correlation factors according to the obtained absolute value, increment value, and growth rate value.

[0085] Specifically, calculate the correlation factor between any two measurement points according to the following calculation formula:

[0086]

[0087] Wherein, w 1 , w 2 and w 3 are weight coefficients, and H is the correlation factor between any two measurement points. The weight coefficient set W is calculated by the improved projection pursuit method, and the calculation result is W = [0.2496, 0.1585, 0.5919].

[0088] Construct a correlation factor set according to the correlation factors between any two measurement points, and the construction formula is:

[0089] x L = {H 1 , H 2 ,,, H m}

[0090] Wherein, x L is the correlation factor set, and H m is the m-th correlation factor. Among them, m = n(n - 1) / 2, n is the number of measurement points, and m represents the number of pairwise combinations between measurement points.

[0091] Furthermore, in step S105, calculating the difference value between the prediction sequence and the real data sequence includes:

[0092] Calculating the difference value between the prediction sequence and the real data sequence using the L2 norm.

[0093] Specifically, by performing real-time sliding segmentation on the newly generated real-time online monitoring data, inputting the data segment into the informer model, and determining anomalies by detecting the difference between the predicted data sequence y and the real data sequence x under the detection window. In the embodiment of the present invention, the L2 norm is used to measure the difference between the predicted data sequence and the real data sequence, and its calculation process is:

[0094] R = R(x, y) = ||x - y|| 2

[0095] The following is the specific process of an abnormal detection method for the internal insulation state of a transformer in the embodiment of the present invention.

[0096] The real data x t collected on the current date is combined with the real data X w-1 of the previous w - 1 days t-w+1 = [x t-w+2 , x t-1 , …, x 1 , H 2 , H m to form a real data sequence X w= [x t-w+1 , x t-w+2 , …, x t , H 1 , H 2 ,,, H m , where w is the time window length. In the experiment, the window length is set to w = 40. The informer prediction model uses X w as the input sequence and predicts the value at the next moment Predicted value and the predicted values at the previous w - 1 moments to form a sequence of prediction data under a fixed window When the real data at time t + 1 is obtained, update the sequence X of real data under a fixed window w = [x t-w+2 , x t-w+3 , …, x t+1 , H 1 , H 2 ,,, H m . Calculate the difference between the sequence X w and For the determination threshold δ, if R ≥ δ, then determine that the latest input data x is abnormal data, otherwise, determine that x t+1 is normal data. t+1

[0097] If multiple consecutive input sequences are judged to be abnormal, it is considered that the internal insulation state of the transformer is abnormal.

[0098] For the method for detecting the abnormal internal insulation state of a transformer according to the embodiment of the present invention, in order to more timely regularize the oil temperature sensor data and more accurately detect the internal insulation state of the transformer according to the sensor data, first analyze the abnormal types of the oil temperature sensor data and perform targeted data cleaning and normalization processing, then through the informer model, input the real data collected in real time to predict the data at future moments. Subsequently, further use the difference between the predicted data sequence and the real data sequence to detect the abnormal insulation state, and can identify the abnormal state in a long-term oil temperature monitoring data, with high detection accuracy.

[0099] The embodiment of the present invention also provides a device for detecting the abnormal internal insulation state of a transformer, as shown in Figure 8 and includes:

[0100] Factor calculation module 801, which is used to calculate the correlation factor set between the data sequences collected at any two different positions inside the transformer; for specific content, refer to the above method embodiment and will not be elaborated here. ​

[0101] The vector construction module 802 is used to construct an input vector based on the real data collected at each measuring point on the current date, the real data collected in the preset number of days before the current date, and the associated factor set; for specific content, refer to the above method embodiments and will not be elaborated here.

[0102] The prediction module 803 is used to input the input vector into the informer model and obtain the predicted values of the data collected at each measuring point on the next day through the informer model; for specific content, refer to the above method embodiments and will not be elaborated here.

[0103] The sequence construction module 804 is used to construct a prediction sequence based on the predicted values of consecutive preset days and the associated factor set, and construct a real data sequence based on the real data actually collected at each measuring point in consecutive preset days and the associated factor set; for specific content, refer to the above method embodiments and will not be elaborated here.

[0104] The anomaly determination module 805 is used to calculate the difference value between the prediction sequence and the real data sequence and determine whether the internal insulation state of the transformer is abnormal based on the difference value. For specific content, refer to the above method embodiments and will not be elaborated here.

[0105] Optionally, the factor calculation module 801 includes:

[0106] The parameter calculation module is used to calculate the absolute value, increment value, and growth rate value between the data sequences collected at the measuring points at any two different positions inside the transformer.

[0107] The factor determination module is used to determine the associated factor set according to the obtained absolute value, increment value, and growth rate value.

[0108] Optionally, the formula for calculating the absolute value between the data sequences collected at the measuring points at any two different positions inside the transformer is:

[0109]

[0110] In the formula, L 1 represents the absolute value, x it and x jt are the data sequences of the measuring points i and j at the t moment respectively, and T is the data acquisition time window.

[0111] Optionally, the formula for calculating the increment value between the data sequences collected at the measuring points at any two different positions inside the transformer is:

[0112]

[0113] In the formula, L 2 represents the increment value, y it and y jtThey are the change amounts of the data sequences of measurement point i and measurement point j at time t relative to time t-1 respectively.

[0114] Optionally, the calculation formula for the growth rate value between the data sequences collected at any two different positions inside the transformer is:

[0115]

[0116] In the formula, L 3 represents the growth rate value, z it and z jt are the change amplitudes of the increment values of the data sequences of measurement point i and measurement point j at time t relative to time t-1 respectively.

[0117] Optionally, the factor determination module 801 is further configured to:

[0118] Calculate the correlation factor of any two measurement points according to the following calculation formula:

[0119]

[0120] In the formula, w 1 、w 2 and w 3 are weight coefficients, and H is the correlation factor of any two measurement points;

[0121] Construct a correlation factor set according to the correlation factors of any two measurement points, and the construction formula is:

[0122] x L ={H 1 ,H 2 ,,,H m}

[0123] In the formula, x L is the correlation factor set, and H m is the mth correlation factor.

[0124] Optionally, the internal insulation state abnormal detection device of the transformer includes:

[0125] A data cleaning module for cleaning the collected real data;

[0126] A normalization module for normalizing the cleaned real data.

[0127] Optionally, the abnormal determination module 805 includes a difference value calculation module for calculating the difference value between the prediction sequence and the real data sequence using the L2 norm.

[0128] An abnormal detection device for the internal insulation state of a transformer provided by an embodiment of the present invention calculates a set of correlation factors between data sequences collected at measurement points at any two different positions inside the transformer; constructs an input vector according to the real data collected at each measurement point on the current date, the real data collected in a preset number of days before the current date, and the set of correlation factors; inputs the input vector into the Informer model, and obtains the predicted values of the data collected at each measurement point on the next day through the Informer model; constructs a prediction sequence based on the predicted values of consecutive preset days and the set of correlation factors, and constructs a real data sequence based on the real data actually collected at each measurement point on consecutive preset days and the set of correlation factors; calculates the difference value between the prediction sequence and the real data sequence, and determines whether the internal insulation state of the transformer is abnormal based on the difference value. Using the Informer model for prediction can reduce the computational complexity, significantly improve the information extraction ability of long sequences, represent the correlation of data at different measurement points through the set of correlation factors, and input it as part of the input vector to synthesize the data of different measurement points to extract the same data trend, so as to make certain corrections to the input data, increase the accuracy of the predicted data, greatly improve the performance of long-term sequence prediction, improve the accuracy of the predicted data, and further increase the accuracy of abnormal data detection.

[0129] An embodiment of the present invention also provides a computer device, as Figure 9 shown. The computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some alternative embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (such as an array of servers, a set of blade servers, or a multi-processor system). Figure 9 Taking one processor 10 as an example in

[0130] The processor 10 can be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above hardware chip can be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above programmable logic device can be a complex programmable logic device, a field programmable gate array, a general array logic, or any combination thereof.

[0131] Among them, the memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiments.

[0132] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0133] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive; the memory 20 may also include a combination of the above types of memories.

[0134] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 may be connected through a bus or other means. Figure 9 Taking the connection through the bus as an example.

[0135] The input device 30 may receive input digital or character information, and generate key signal inputs related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (such as an LED), and a tactile feedback device (such as a vibration motor), etc. The above display device includes but is not limited to a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device may be a touch screen.

[0136] Embodiments of the present invention also provide a computer-readable storage medium. The methods according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code that is originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the methods described herein can be stored in such software processes on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0137] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for detecting abnormal internal insulation state of a transformer, characterized in that, it includes: Calculating the set of correlation factors between the data sequences collected by the measuring points at any two different positions inside the transformer; Constructing an input vector based on the real data collected by each measuring point on the current date, the real data collected in a preset number of days before the current date, and the set of correlation factors; Inputting the input vector into the Informer model, and obtaining the predicted values of the data collected by each measuring point on the next day through the Informer model; Constructing a prediction sequence based on the predicted values of the continuous preset number of days and the set of correlation factors, and constructing a real data sequence based on the real data actually collected by each measuring point in the continuous preset number of days and the set of correlation factors; Calculating the difference value between the prediction sequence and the real data sequence, and judging whether the internal insulation state of the transformer is abnormal based on the difference value.

2. The method according to claim 1, characterized in that, The calculating the set of correlation factors between the data sequences collected by the measuring points at any two different positions inside the transformer includes: Calculating the absolute value, increment value and growth rate value between the data sequences collected by the measuring points at any two different positions inside the transformer; Determining the set of correlation factors according to the obtained absolute value, increment value and growth rate value.

3. The method according to claim 2, characterized in that, The formula for calculating the absolute value between the data sequences collected by the measuring points at any two different positions inside the transformer is: Where L 1 represents the absolute value, x it and x jt are the data sequences of measurement point i and measurement point j at time t respectively, and T is the data acquisition time window.

4. The method according to claim 3, characterized in that, The formula for calculating the increment value between the data sequences collected by the measuring points at any two different positions inside the transformer is: where L 2 represents the incremental value, and y it and y jt are the change amounts of the data sequences of measurement point i and measurement point j at time t relative to time t - 1, respectively.

5. The method according to claim 4, characterized in that, The formula for calculating the growth rate value between the data sequences collected by the measuring points at any two different positions inside the transformer is: where L 3 represents the growth rate value, and z it and z jt are the change amplitudes of the increment values of the data sequences at measurement point i and measurement point j at time t relative to time t - 1, respectively.

6. The method according to claim 5, characterized in that, Determining the set of correlation factors according to the obtained absolute value, increment value and growth rate value includes: Calculating the correlation factor between any two measuring points according to the following calculation formula: where w 1 , w 2 and w 3 are weight coefficients, and H is the correlation factor between any two measurement points; Constructing a set of correlation factors according to the correlation factors between any two measuring points, and the construction formula is: x L = {H 1 , H 2 ,,, H m} where x L is the set of correlation factors, and H m is the m-th correlation factor.

7. The method according to claim 1, characterized in that, Before constructing the input vector based on the real data collected by each measuring point on the current date, the real data collected in a preset number of days before the current date, and the set of correlation factors, it includes: Performing data cleaning on the collected real data; Performing normalization processing on the cleaned real data.

8. The method according to claim 1, characterized in that, Calculating the difference value between the prediction sequence and the real data sequence includes: Calculating the difference value between the prediction sequence and the real data sequence using the L2 norm.

9. A device for detecting abnormal internal insulation state of a transformer, characterized in that, it includes: A factor calculation module for calculating the set of correlation factors between the data sequences collected by the measuring points at any two different positions inside the transformer; A vector construction module, configured to construct an input vector based on the real data collected by each measurement point on the current date, the real data collected in a preset number of days before the current date, and the associated factor set; A prediction module, configured to input the input vector into an informer model, and obtain predicted values of the data collected by each measurement point on the next day through the informer model; A sequence construction module, configured to construct a prediction sequence based on the predicted values of consecutive preset days and the associated factor set, and construct a real data sequence based on the real data actually collected by each measurement point on consecutive preset days and the associated factor set; An anomaly determination module, configured to calculate a difference value between the prediction sequence and the real data sequence, and determine whether the internal insulation state of the transformer is abnormal based on the difference value.

10. A computer device, characterized in that it includes: A memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the method for detecting abnormal internal insulation state of a transformer according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that computer instructions are stored on the computer-readable storage medium, and the computer instructions are used to cause a computer to execute the method for detecting abnormal internal insulation state of a transformer according to any one of claims 1 to 8.

Images