Oil-immersed transformer condition assessment method and device based on graph convolutional neural network

Abstract

The present invention discloses a method and device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network, which relates to the technical field of power systems and smart grids, and is mainly intended to solve the problem of low efficiency of existing oil-immersed transformer state monitoring. It mainly includes obtaining the time series data of multiple influencing factors of the target transformer associated with preset operating state evaluation parameters within a preset time period; according to the correlation between the various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data; the unweighted undirected graph structure time series data is predicted and processed using a pre-trained influencing factor prediction model to obtain the prediction data of different influencing factors at the next moment; and the state evaluation result is determined based on the prediction data of different influencing factors. Thereby improving the evaluation efficiency and the timeliness of transformer monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems and smart grids, and in particular to a method and device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network. Background Art

[0002] With the development of modern power systems and the continuous expansion of power grids, transformers, as key equipment in power systems, are directly related to the stability and reliability of power grids in terms of their operating status. Transformer failures not only lead to power outages, but may also cause serious economic losses and safety hazards. Therefore, it is of great practical significance to effectively monitor the operating status of transformers to detect potential faults in a timely manner and take preventive measures.

[0003] Traditional transformer condition monitoring methods mainly rely on periodic manual detection and simple threshold alarms. This method has problems such as long detection cycle, slow response speed, and low fault identification accuracy. As a result, the efficiency of oil-immersed transformer condition monitoring is low and it is difficult to meet the high requirements of modern power systems for transformer condition monitoring. Summary of the invention

[0004] In view of this, the present invention provides a method and device for oil-immersed transformer state assessment based on graph convolutional neural network, the main purpose of which is to solve the problem of low efficiency of existing oil-immersed transformer state monitoring.

[0005] According to one aspect of the present invention, a method for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network is provided, comprising:

[0006] Acquire time series data of multiple influencing factors associated with preset operating status evaluation parameters of the target transformer within a preset time period, wherein the operating status evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters;

[0007] According to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data;

[0008] Using a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment;

[0009] A plurality of target operating state evaluation parameters are determined according to the prediction data of different influencing factors, and a state evaluation result is determined according to the target operating state evaluation parameters.

[0010] Furthermore, according to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into time series data with an unweighted undirected graph structure, including:

[0011] Take each influencing factor as a node, determine the edge between nodes according to the correlation between the influencing factors corresponding to each two nodes, and construct an unweighted undirected graph;

[0012] According to the adjacency matrix of the unweighted undirected graph, the influencing factors corresponding to each time point in the time series data are sequentially introduced into the unweighted undirected graph to obtain the unweighted undirected graph structure time series data.

[0013] Furthermore, before using the pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data, the method also includes:

[0014] Acquire historical time series data of influencing factors, use the historical data of the influencing factors in a target period as target input, and use the historical data of the influencing factors at the next moment of the target period as prediction labels to construct training samples, thereby obtaining a training sample set;

[0015] According to a preset time axis division ratio, the training sample set is divided into a training set and a test set from a time dimension;

[0016] Using the training set to train the graph convolutional neural network model to obtain an evaluation model;

[0017] The test set is used to perform model performance testing and parameter optimization on the evaluation model to obtain a pre-trained influencing factor prediction model.

[0018] Furthermore, the dynamic electrical quantity dimension is represented by the transformer transmission efficiency, the oil temperature dimension is represented by the rate of change of the oil temperature to the ambient temperature, the influencing factors include the oil temperature, the ambient temperature, the high-voltage side active power and the low-voltage side active power, and the multiple target operating state evaluation parameters are determined based on the predicted data of different influencing factors, including:

[0019] According to the predicted value of active power on the high voltage side and the predicted value of active power on the low voltage side, the transformer transmission efficiency is calculated, and the transformer transmission efficiency is used as the first target operating state evaluation parameter, wherein the calculation formula is expressed as Wherein, η represents the predicted value of transformer transmission power, P1 represents the predicted value of active power on the high voltage side, and P2 represents the predicted value of active power on the low voltage side;

[0020] According to the predicted value of oil temperature and the predicted value of ambient temperature, the change rate of oil temperature to ambient temperature is calculated, and the change rate of oil temperature to ambient temperature is used as the second target operating state evaluation parameter, wherein the calculation formula is expressed as: Wherein, λ represents the predicted value of the rate of change, Δθ represents the predicted value of the top oil temperature change during a preset period, and ΔT represents the predicted value of the external environment temperature change during a preset period.

[0021] Furthermore, the influencing factors also include high-voltage side current and low-voltage side current. Before determining a plurality of target operating state evaluation parameters based on the prediction data of different influencing factors, the method further includes:

[0022] The high-voltage side active power verification value is calculated based on the high-voltage side current prediction value and the current high-voltage side voltage value;

[0023] The low-voltage side active power verification value is calculated based on the low-voltage side current prediction value and the current low-voltage side voltage value;

[0024] The high-voltage side active power prediction value is verified according to the high-voltage side active power verification value, and the high-voltage side active power prediction value is verified according to the low-voltage side active power verification value. If the result of any of the verifications fails, return to the step of obtaining the operating data of the target transformer within the preset time period.

[0025] Furthermore, the vibration signal dimension parameter is expressed as frequency concentration, and the method further includes:

[0026] Perform spectrum analysis on the vibration signal and separate the amplitude of the nth harmonic with a fundamental frequency of 100 Hz;

[0027] The signal frequency concentration is calculated according to the amplitude of the nth harmonic with the fundamental frequency of 100 Hz, and the signal frequency concentration is used as the third target operating state evaluation parameter; wherein the calculation formula of the signal frequency concentration is expressed as: DFC stands for frequency concentration, a 100n Indicates the amplitude of the nth harmonic with a fundamental frequency of 100 Hz, a 100 Indicates the vibration amplitude of the component with a fundamental frequency of 100 Hz.

[0028] Furthermore, the status assessment result includes normal, caution, abnormal and severe, and the status assessment result is determined according to the target operation status assessment parameter, including:

[0029] Calculate the difference between each target operating status assessment parameter and the corresponding parameter limit and the fluctuation within a preset time interval;

[0030] If the difference value of any target operating status evaluation parameter represents an abnormal difference direction, and the absolute value of the difference value is greater than the first difference threshold, then the status evaluation result is determined to be severe;

[0031] If the fluctuation amount of any target operating state evaluation parameter is greater than the fluctuation threshold, and the absolute value of the difference value is less than the second difference threshold, then determining that the state evaluation result is abnormal;

[0032] If the difference value of any target operating status evaluation parameter represents a normal difference direction, and the difference value is less than the second difference threshold, determining that the status evaluation result is warning;

[0033] If the fluctuation amount of any target operating status evaluation parameter is less than or equal to the fluctuation threshold, and the difference value represents a normal difference direction, then the status evaluation result is determined to be normal.

[0034] According to another aspect of the present invention, a device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network is provided, comprising:

[0035] An acquisition module, used to acquire time series data of multiple influencing factors associated with preset operating status evaluation parameters of a target transformer within a preset time period, wherein the operating status evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters;

[0036] A conversion module, for converting the time series data into time series data with an unweighted undirected graph structure according to the correlation between various influencing factors and taking the data at each time point as a graph state;

[0037] A processing module, used to use a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment;

[0038] A determination module is used to determine a plurality of target operating status evaluation parameters according to the prediction data of different influencing factors, and determine a status evaluation result according to the target operating status evaluation parameters.

[0039] According to another aspect of the present invention, a storage medium is provided, in which at least one executable instruction is stored, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned oil-immersed transformer state assessment method based on graph convolutional neural network.

[0040] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0041] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned oil-immersed transformer state assessment method based on graph convolutional neural network.

[0042] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0043] The present invention provides a method and device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network. The embodiment of the present invention obtains the time series data of multiple influencing factors of the target transformer in a preset time period and associates the preset operating state evaluation parameters, and the operating state evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters and vibration signal dimension parameters; according to the correlation between the various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data; the unweighted undirected graph structure time series data is predicted and processed using a pre-trained influencing factor prediction model to obtain the prediction data of different influencing factors at the next moment; multiple target operating state evaluation parameters are determined based on the prediction data of different influencing factors, and the state evaluation results are determined based on the target operating state evaluation parameters. The time of transformer operating state evaluation is greatly reduced, the timeliness of monitoring is improved, and at the same time, the accuracy of transformer monitoring is ensured, thereby greatly improving the efficiency of transformer state evaluation.

[0044] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0046] Figure 1 A flow chart of a method for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network provided by an embodiment of the present invention is shown;

[0047] Figure 2 Another flow chart of an oil-immersed transformer state assessment method based on a graph convolutional neural network provided by an embodiment of the present invention is shown;

[0048] Figure 3 A schematic diagram of an unweighted undirected graph provided by an embodiment of the present invention is shown;

[0049] Figure 4 A schematic diagram of training set division provided by an embodiment of the present invention is shown;

[0050] Figure 5 A training set partition result code provided by an embodiment of the present invention is shown;

[0051] Figure 6 A comparison diagram of a predicted value and an actual value of an influencing factor provided by an embodiment of the present invention is shown;

[0052] Figure 7 A block diagram of a device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network provided by an embodiment of the present invention is shown;

[0053] Figure 8 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0054] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0055] Aiming at the problem of low efficiency of existing oil-immersed transformer condition monitoring. The embodiment of the present invention provides a method for evaluating the condition of an oil-immersed transformer based on a graph convolutional neural network, such as Figure 1 As shown, the method includes:

[0056] 101. Obtain time series data of multiple influencing factors associated with preset operating status evaluation parameters of a target transformer within a preset time period.

[0057] In an embodiment of the present invention, the target transformer is an oil-immersed transformer that needs to be monitored in the power system. The operating status evaluation parameters are index parameters determined in advance according to the working characteristics of large oil-immersed transformers. The parameters used to evaluate the operating status of the transformer may include dynamic electrical quantity dimension parameters, oil temperature dimension parameters and vibration signal dimension parameters. The influencing factors are operating data items arranged in time series corresponding to each operating status evaluation parameter and capable of affecting the numerical value of the operating status evaluation parameter, and may include dynamic electrical quantity data, oil temperature data and vibration signals. Among them, the dynamic electrical quantity data and oil temperature data are both from the substation background data monitoring and storage system, and this system completes sampling every 5 minutes. The vibration signal acquisition selects a test scheme with a three-phase median value of 2000Hz. The three-phase median value of 2000Hz refers to the arrangement of sensors in the middle position of the three phases of the low-voltage side box of the transformer, and the sampling frequency is 2000Hz.

[0058] 102. According to the correlation between various influencing factors, the data at each time point is taken as a graph state, and the time series data is converted into unweighted undirected graph structure time series data.

[0059] In an embodiment of the present invention, an unweighted undirected graph is used to represent the relationship between various influencing factors. The transformer operating status evaluation parameters are selected from three aspects: electrical, mechanical and thermal. Since the three influencing factors are related but have no clear numerical relationship, an unweighted undirected graph can be used to fuzzily represent them. In addition, since it is necessary to predict the value at the next moment based on the current value of each influencing factor. Therefore, it is necessary to convert the time series data into time series data with an unweighted undirected graph structure, that is, each influencing factor is used as a node of the unweighted undirected graph, the relationship between each influencing factor is represented by the edge between the nodes, and the influencing factor feature corresponding to each time point in the time series data is used as a graph state of the unweighted undirected graph, thereby obtaining the unweighted undirected graph structure time series data.

[0060] 103. Use a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment.

[0061] In the embodiment of the present invention, the values ​​of the influencing factors of the transformer are predicted by using the pre-trained influencing factor prediction model to obtain the prediction data of each influencing factor at the next moment. Taking an influencing factor as an example, the historical moment data of the influencing factor [X t-s ,…,X t-2 ,X t-1 ] to predict the influencing factors at the next moment t, namely: X t =F([X t-s ,…,X t-2 ,X t-1 ])(1); where X is the feature vector of the influencing factor; F is the function of the pre-trained influencing factor prediction model, and s is the number of historical data.

[0062] It should be noted that by associating the influencing factors corresponding to each operating status evaluation parameter through an unweighted undirected graph and integrating the time series attributes into the unweighted undirected graph, in the prediction process of the model, we can pay attention to both the correlation between the various influencing factors and the changing characteristics of each influencing factor over time, thereby achieving efficient and accurate prediction of multiple influencing factors.

[0063] 104. Determine a plurality of target operating status evaluation parameters based on the prediction data of different influencing factors, and determine a status evaluation result based on the target operating status evaluation parameters.

[0064] In an embodiment of the present invention, the target operating state evaluation parameter is an operating state evaluation parameter required to evaluate the operating state of the target transformer. It includes operating state evaluation parameters that can be directly calculated through the predicted data of different influencing factors, and also includes operating state evaluation parameters that need to be calculated based on actual operating data. The parameter dimension of the target operating state evaluation parameter is the same as the dimension of the preset operating state evaluation parameter, that is, it is necessary to comprehensively evaluate the operating state of the target transformer using the target operating state evaluation parameters of multiple dimensions such as electrical, thermal, and mechanical to obtain a state evaluation result.

[0065] In an embodiment of the present invention, for further explanation and limitation, as Figure 2 As shown, according to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data, including:

[0066] 201. Take each influencing factor as a node, determine the edge between the nodes according to the correlation between the influencing factors corresponding to each two nodes, and construct an unweighted undirected graph;

[0067] 202. According to the adjacency matrix of the unweighted undirected graph, the influencing factors corresponding to each time point in the time series data are sequentially imported into the unweighted undirected graph to obtain the unweighted undirected graph structure time series data.

[0068] In the embodiment of the present invention, in the unweighted undirected graph, each node represents an influencing factor, and the node feature is a characteristic index of the influencing factor. If there is an obvious influencing relationship between two factors, the two nodes are neighbor nodes, that is, they are connected by a connecting line, and are represented as non-zero values ​​in the adjacency matrix. Figure 3 As shown, it is an unweighted undirected graph, and the nodes in the graph correspond to six influencing factors: high-voltage side current, low-voltage side current, high-voltage side active power, low-voltage side active power, oil temperature and ambient temperature. The relationship between nodes in the unweighted undirected graph is represented by 1 when there is a connection between nodes and 0 when there is no connection, and an adjacency matrix is ​​obtained. As shown in Table 1, it is the adjacency matrix corresponding to this unweighted undirected graph. The influencing factors corresponding to each time point in the time series data are imported into the unweighted undirected graph according to the time sequence and the relationship between the influencing factors represented by the adjacency matrix, so that the unweighted undirected graph presents different states in time sequence, thereby obtaining the unweighted undirected graph structure time series data.

[0069] Table 1:

[0070] High current Low current Gao Yougong Low active power Oil temperature Ambient temperature High current 0 1 1 0 0 0 Low current 1 0 0 1 0 0 Gao Yougong 1 0 0 1 1 0 Low active power 0 1 1 0 1 0 Oil temperature 0 0 1 1 0 1 Ambient temperature 0 0 0 0 1 0

[0071] In one embodiment of the present invention, for further explanation and limitation, before using the pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data, the method further includes:

[0072] Acquire historical time series data of influencing factors, use the historical data of the influencing factors in a target period as target input, and use the historical data of the influencing factors at the next moment of the target period as prediction labels to construct training samples, thereby obtaining a training sample set;

[0073] According to a preset time axis division ratio, the training sample set is divided into a training set and a test set from a time dimension;

[0074] Using the training set to train the graph convolutional neural network model to obtain an evaluation model;

[0075] The test set is used to perform model performance testing and parameter optimization on the evaluation model to obtain a pre-trained influencing factor prediction model.

[0076] In the embodiment of the present invention, since the historical time series data is a series of data arranged in time, that is, these data are distributed on the time axis. In order to construct a training sample, data of a length of H moments on the time axis is used to predict the data at the H+1 moment. Among them, the data of the first H moments can be regarded as the input x in the sample, and the data of the H+1th moment can be regarded as the label y of the sample. (x, y) constitutes a training data pair, that is, a data sample. According to the above-mentioned construction process of a single training sample, multiple training samples are constructed to obtain a sample training set. After determining the training sample set, the data set is segmented, and a reasonable segmentation ratio (preset time axis division ratio) is preset, and the data distributed on the time axis is segmented to obtain a training set and a test set that meet the testing requirements of the model training set. As Figure 4 As shown, the complete distribution length of the historical sequence data on the time axis is the time axis length T, the target period length is H, the training set length is L1, and the test set length is L2. Let the sample subscript be idx, and reserve a target period length H. Then the subscript range of the training set is [H, L1-1+H], and the test set subscript is [0, L2-1]. The relationship between the time axis length T, the target period length H, the training set length L1, and the test set length L2 satisfies the following formula: T = H + L1 + L2 (2). Among them, the preset time axis division ratio can be 3:1, and can also be customized according to actual application requirements, which is not specifically limited in the embodiments of the present invention.

[0077] In the actual test with 1152 pairs of original data sets, the duration (number of days) is divided according to the ratio of 3:1 to obtain the training set and the test set. Select 6 moments as the historical data length, that is, select 30 minutes of data as the historical data (data within the target duration), and predict the data of each influencing factor at the next 5 minutes. In theory, 1152÷4×3-6=858 training samples can be obtained. Divide the training set and the test set according to this division criterion, and verify the accuracy of the division. Figure 5 The following is the result code of data set partitioning verification. 858 refers to the length of the obtained training set, which is the same as the expected number of partitioned training samples, and the dimensions of the training set and the test set are also the same as the expected results, proving that the transformer state factor time series data modeling is effective.

[0078] In the embodiment of the present invention, the graph convolutional neural network is a graph convolutional neural network based on the spatial domain, and its model expression can be expressed as: Where Y represents the predicted output, X is the model input, and W is the model training weight. Obtained by the following formula: In the formula, I is the unit matrix of size 1, and D is the degree matrix. Based on the Python algorithm, the prediction is performed to obtain the sample prediction results. The model is trained based on the loss function constructed based on the sample prediction results and the prediction labels to obtain the evaluation model. The evaluation model is tested using the test set to obtain a pre-trained influencing factor prediction model whose model performance meets expectations. Figure 6 As shown in FIG. 1 , it is a comparison diagram of the predicted data and the actual value of each influencing factor. Based on this method, a relatively accurate predicted value of the influencing factor can be obtained. The high in the figure refers to the high pressure side, and the low refers to the low pressure side.

[0079] In one embodiment of the present invention, for further explanation and limitation, the multiple target operating state evaluation parameters are determined based on the prediction data of different influencing factors, including:

[0080] According to the predicted value of active power on the high voltage side and the predicted value of active power on the low voltage side, the transformer transmission efficiency is calculated, and the transformer transmission efficiency is used as the first target operating state evaluation parameter;

[0081] The change rate of the oil temperature to the ambient temperature is calculated based on the predicted value of the oil temperature and the predicted value of the ambient temperature, and the change rate of the oil temperature to the ambient temperature is used as the second target operating state evaluation parameter.

[0082] In the embodiment of the present invention, the dynamic electrical quantity dimension is represented by the transformer transmission efficiency, and the corresponding influencing factors are the high-voltage side active power and the low-voltage side active power. The calculation result corresponds to the first target operating state evaluation parameter. The calculation formula is expressed as: Among them, η represents the predicted value of transformer transmission power, P1 represents the predicted value of high-voltage side active power, and P2 represents the predicted value of low-voltage side active power. The oil temperature dimension is represented by the rate of change of oil temperature to ambient temperature, and the corresponding influencing factors are oil temperature and ambient temperature. The calculation result corresponds to the second target operating state evaluation parameter. Its calculation formula is expressed as Wherein, λ represents the predicted value of the rate of change, Δθ represents the predicted value of the top oil temperature change during a preset period, and ΔT represents the predicted value of the external environment temperature change during a preset period.

[0083] It should be noted that due to the large data differences between oil temperature, active power values ​​on the high and low voltage sides, current values ​​on the high and low voltage sides, and ambient temperature, if they are directly substituted into the neural network model, the prediction accuracy will deteriorate, resulting in inaccurate final state assessment results. Therefore, in order to ensure that the graph convolutional neural network model can effectively identify the changing relationship between different influencing factors, the time series data of the influencing factors needs to be normalized before entering the model. Specifically, data normalization is performed based on the following formula:

[0084]

[0085] In the formula, X′ represents the data result after normalization, X represents the data before normalization, and X min Refers to the minimum value of the corresponding data in the entire data set, X max Refers to the maximum value of the corresponding data in the entire data set. After obtaining the predicted values ​​of different influencing factors, the predicted values ​​are denormalized to calculate the target operating state evaluation parameters based on the denormalized predicted values.

[0086] In one embodiment of the present invention, for further explanation and limitation,

[0087] Furthermore, before determining a plurality of target operating state evaluation parameters based on the prediction data of different influencing factors, the method further includes:

[0088] The high-voltage side active power verification value is calculated based on the high-voltage side current prediction value and the current high-voltage side voltage value;

[0089] The low-voltage side active power verification value is calculated based on the low-voltage side current prediction value and the current low-voltage side voltage value;

[0090] The high-voltage side active power prediction value is verified according to the high-voltage side active power verification value, and the high-voltage side active power prediction value is verified according to the low-voltage side active power verification value. If the result of any of the verifications fails, return to the step of obtaining the operating data of the target transformer within the preset time period.

[0091] In an embodiment of the present invention, in order to improve the accuracy of the calculation of the target operating state evaluation parameters, cross-validation is performed based on the correlation between multiple influencing factors, and the high-voltage side current and the low-voltage side current are added to the influencing factors. Although the high-voltage side current and the low-voltage side current cannot directly calculate the transformer transmission efficiency. However, the product of the high-voltage side current and the actual voltage value can be used to obtain the high-voltage side active power. The predicted value of the high-voltage side active power is verified based on the predicted high-voltage side current and the current actual high-voltage side voltage value as the verification value of the high-voltage side active power. If the verification value is significantly different from the predicted value, it indicates that the prediction accuracy is low, and data collection and prediction can be performed again to improve the accuracy of the prediction, thereby improving the accuracy of the target operating state evaluation parameters. Similarly, the low-voltage side can also be verified in the same way.

[0092] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0093] Perform spectrum analysis on the vibration signal and separate the amplitude of the nth harmonic with a fundamental frequency of 100 Hz;

[0094] The signal frequency concentration is calculated according to the amplitude of the nth harmonic with the fundamental frequency of 100 Hz, and the signal frequency concentration is used as the third target operating state evaluation parameter.

[0095] In the embodiment of the present invention, when the transformer is mechanically stable, according to the transformer vibration mechanism, the signal frequency is concentrated at 100Hz, and the DFC is close to 1; on the contrary, the transformer core is loose or the winding is deformed, which increases the high-frequency component in the vibration signal, disperses the frequency distribution, and reduces the DFC. That is, the vibration signal of the abnormal state of the transformer is manifested as a reduction in the 100Hz component and an increase in the vibration amplitude. Therefore, the frequency concentration is selected to represent the vibration signal dimension parameter. Among them, the calculation formula of the signal frequency concentration is expressed as: Where DFC represents frequency concentration, a 100n Indicates the amplitude of the nth harmonic with a fundamental frequency of 100 Hz, a 100 Indicates the vibration amplitude of the component with a fundamental frequency of 100 Hz.

[0096] In one embodiment of the present invention, for further explanation and limitation, determining the status evaluation result according to the target operating status evaluation parameter includes:

[0097] Calculate the difference between each target operating status assessment parameter and the corresponding parameter limit and the fluctuation within a preset time interval;

[0098] If the difference value of any target operating status evaluation parameter represents an abnormal difference direction, and the absolute value of the difference value is greater than the first difference threshold, then the status evaluation result is determined to be severe;

[0099] If the fluctuation amount of any target operating state evaluation parameter is greater than the fluctuation threshold, and the absolute value of the difference value is less than the second difference threshold, then determining that the state evaluation result is abnormal;

[0100] If the difference value of any target operating status evaluation parameter represents a normal difference direction, and the difference value is less than the second difference threshold, determining that the status evaluation result is warning;

[0101] If the fluctuation amount of any target operating status evaluation parameter is less than or equal to the fluctuation threshold, and the difference value represents a normal difference direction, then the status evaluation result is determined to be normal.

[0102] In the embodiment of the present invention, the status assessment results include normal, attention, abnormal and severe. It should be noted that the parameter limit can be configured as a transformer transmission efficiency of 95%, a frequency concentration of 0.7, and an oil temperature change rate of 4 to the ambient temperature. The specific reasons are as follows: Since the transmission efficiency of a large transformer under normal operation is above 95%, 95% can be used as the standard limit of the transmission efficiency. When the transmission efficiency is lower than 95%, it can be determined that the transmission efficiency is abnormal. The frequency concentration of the transformer during normal operation is usually around 0.8. 0.7 can be set as the parameter limit of the transformer frequency concentration. When the frequency concentration is lower than 0.7, it can be determined that the vibration signal is abnormal. After statistical calculation of the normally operating transformer, the change rate of its oil temperature to the ambient temperature is less than 4. 4 can be set as the parameter limit of the oil temperature change rate to the ambient temperature. When the value is greater than 4, it is determined that the transformer oil temperature is abnormal. Of course, the parameter limits corresponding to different operating status assessment parameters can also be customized according to the actual application situation, and the embodiment of the present invention does not make specific restrictions.

[0103] The difference value is the difference between the target operating status evaluation parameter and the corresponding parameter limit. This difference value can be positive, negative or 0. The positive and negative difference values ​​have different meanings for different target operating status evaluation parameters. For example, the transmission efficiency of a large transformer under normal working conditions is above 95%. Assume that the parameter limit corresponding to the transmission efficiency is 95%. If the difference value is positive or 0, it represents the normal difference direction; if the difference value is negative, that is, the transmission efficiency is less than the parameter limit, it represents the abnormal difference direction. For another example, after statistical calculation of the normally operating transformer, the rate of change of its oil temperature to the ambient temperature is less than 4. Assume that 4 is the parameter limit of the rate of change of the oil temperature to the ambient temperature. If the difference value is positive or 0, it represents the abnormal difference direction; if the difference value is negative, it represents the normal difference direction. If the difference value represents the abnormal difference direction, and the absolute value of the difference value is large (greater than the first difference threshold), the state assessment result is determined to be serious. If the difference value of one or more target operating status evaluation parameters represents the normal difference direction, and the absolute value of the difference value is small (smaller than the second difference threshold), indicating that the target operating status evaluation parameter is close to the parameter limit, it is necessary to pay close attention, and its status evaluation result is attention. The fluctuation amount within the preset time interval is used to characterize the stability of the target operating status evaluation parameter. If the fluctuation amount is large, and the parameter is close to the parameter limit or slightly exceeds the parameter limit (the absolute value of the difference value is small), it can be determined that the operating state is abnormal. Among them, the first difference threshold is greater than the second difference threshold, and the specific values ​​of the first difference threshold, the second difference threshold and the fluctuation amount can be customized according to the actual application scenario, and the embodiment of the present invention does not make specific limitations.

[0104] In an application example, the predicted data of various influencing factors at a certain moment are shown in Table 2. According to the power data in the table and formula (4), the transformer transmission efficiency can be calculated as: Therefore, it can be determined that the transformer transmission efficiency at this time period is normal. According to the temperature data set formula (5) in the table, the change rate of the transformer oil temperature to the ambient temperature can be calculated as: Therefore, it can be determined that the oil temperature in the time period at this moment is normal with respect to the rate of change of the ambient temperature.

[0105] Table 2:

[0106] Influencing factors Prediction data Oil temperature, °C 2.270128663 Active power value on high voltage side, kW 5.97044656 Active power value on low voltage side, kW -5.82720525 High voltage side current value, A 28.76943327 Current value on low voltage side, A 327.5503584 Ambient temperature, °C 6.121317351

[0107] The present invention provides a state assessment method for oil-immersed transformers based on graph convolutional neural networks. The embodiment of the present invention obtains the time series data of multiple influencing factors of the target transformer in a preset time period and associates the preset operating state assessment parameters, and the operating state assessment parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters and vibration signal dimension parameters; according to the correlation between the various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data; the unweighted undirected graph structure time series data is predicted and processed using a pre-trained influencing factor prediction model to obtain the prediction data of different influencing factors at the next moment; multiple target operating state assessment parameters are determined based on the prediction data of different influencing factors, and the state assessment result is determined based on the target operating state assessment parameters. The time of transformer operating state assessment is greatly reduced, the timeliness of monitoring is improved, and at the same time, the accuracy of transformer monitoring is ensured, thereby greatly improving the efficiency of transformer state assessment.

[0108] Furthermore, as a response to the above Figure 1 The implementation of the method shown in the figure, the embodiment of the present invention provides an oil-immersed transformer state assessment device based on a graph convolutional neural network, such as Figure 7 As shown, the device comprises:

[0109] An acquisition module 31 is used to acquire time series data of multiple influencing factors associated with preset operating state evaluation parameters of a target transformer within a preset time period, wherein the operating state evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters;

[0110] A conversion module 32 is used to convert the time series data into time series data of an unweighted undirected graph structure according to the correlation between various influencing factors and taking the data at each time point as a graph state;

[0111] The processing module 33 is used to use the pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain the prediction data of different influencing factors at the next moment;

[0112] The determination module 34 is used to determine a plurality of target operating state evaluation parameters according to the prediction data of different influencing factors, and determine a state evaluation result according to the target operating state evaluation parameters.

[0113] Furthermore, the conversion module 32 includes:

[0114] A construction unit is used to take each influencing factor as a node, determine the edge between the nodes according to the correlation between the influencing factors corresponding to each two nodes, and construct an unweighted undirected graph;

[0115] The import unit is used to import the influencing factors corresponding to each time point in the time series data into the unweighted undirected graph in sequence according to the adjacency matrix of the unweighted undirected graph, so as to obtain the unweighted undirected graph structure time series data.

[0116] Furthermore, the device also includes:

[0117] A construction module is used to obtain historical time series data of influencing factors, take the historical data of the influencing factors in a target period as target input, and take the historical data of the influencing factors at the next moment of the target period as prediction labels to construct training samples, thereby obtaining a training sample set;

[0118] A division module, used to divide the training sample set into a training set and a test set from a time dimension according to a preset time axis division ratio;

[0119] A training module, used to train the graph convolutional neural network model using the training set to obtain an evaluation model;

[0120] The testing module is used to use the test set to perform model performance testing and parameter optimization on the evaluation model to obtain a pre-trained influencing factor prediction model.

[0121] Furthermore, the determination module 34 includes:

[0122] The first calculation unit is used to calculate the transformer transmission efficiency according to the high-voltage side active power prediction value and the low-voltage side active power prediction value, and use the transformer transmission efficiency as the first target operating state evaluation parameter, wherein the calculation formula is expressed as Wherein, η represents the predicted value of transformer transmission power, P1 represents the predicted value of active power on the high voltage side, and P2 represents the predicted value of active power on the low voltage side;

[0123] The second calculation unit is used to calculate the change rate of the oil temperature to the ambient temperature based on the predicted value of the oil temperature and the predicted value of the ambient temperature, and use the change rate of the oil temperature to the ambient temperature as the second target operating state evaluation parameter, wherein the calculation formula is expressed as follows: Wherein, λ represents the predicted value of the rate of change, Δθ represents the predicted value of the top oil temperature change during a preset period, and ΔT represents the predicted value of the external environment temperature change during a preset period.

[0124] Furthermore, the device also includes:

[0125] The first calculation module is used to calculate the high-voltage side active power verification value according to the high-voltage side current prediction value and the current high-voltage side voltage value;

[0126] The second calculation module is used to calculate the low-voltage side active power verification value according to the low-voltage side current prediction value and the current low-voltage side voltage value;

[0127] A verification module is used to verify the high-voltage side active power prediction value according to the high-voltage side active power verification value, and to verify the high-voltage side active power prediction value according to the low-voltage side active power verification value. If the result of any of the verifications fails, return to the step of obtaining the operating data of the target transformer within a preset time period.

[0128] Furthermore, the confirmation module 34 further includes:

[0129] An analysis unit, used for performing spectrum analysis on the vibration signal and separating the amplitude of the nth harmonic with a fundamental frequency of 100 Hz;

[0130] The third calculation unit is used to calculate the signal frequency concentration according to the amplitude of the nth harmonic with the fundamental frequency of 100 Hz, and use the signal frequency concentration as the third target operating state evaluation parameter; wherein the calculation formula of the signal frequency concentration is expressed as: DFC stands for frequency concentration, a 100n Indicates the amplitude of the nth harmonic with a fundamental frequency of 100 Hz, a 100 Indicates the vibration amplitude of the component with a fundamental frequency of 100 Hz.

[0131] Furthermore, the determining module 34 further includes:

[0132] A fourth calculation unit, used to calculate the difference between each target operating state evaluation parameter and the corresponding parameter limit value and the fluctuation amount within a preset time interval;

[0133] a first determining unit, configured to determine that a status assessment result is severe if a difference value of any target operating status assessment parameter represents an abnormal difference direction and an absolute value of the difference value is greater than a first difference threshold;

[0134] a second determining unit, configured to determine that a state evaluation result is abnormal if the fluctuation amount of any target operating state evaluation parameter is greater than a fluctuation threshold and the absolute value of the difference value is less than a second difference threshold;

[0135] a third determining unit, configured to determine that a status assessment result is a warning if the difference value of any target operating status assessment parameter represents a normal difference direction and the difference value is less than the second difference threshold;

[0136] The fourth determining unit is used to determine that the status assessment result is normal if the fluctuation amount of any target operating status assessment parameter is less than or equal to the fluctuation threshold and the difference value represents a normal difference direction.

[0137] The present invention provides an oil-immersed transformer state assessment device based on a graph convolutional neural network. The embodiment of the present invention obtains the time series data of multiple influencing factors of the target transformer in a preset time period and associates the preset operation state assessment parameters, and the operation state assessment parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters and vibration signal dimension parameters; according to the correlation between the various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data; the unweighted undirected graph structure time series data is predicted and processed using a pre-trained influencing factor prediction model to obtain the prediction data of different influencing factors at the next moment; multiple target operation state assessment parameters are determined based on the prediction data of different influencing factors, and the state assessment result is determined based on the target operation state assessment parameters. The time of transformer operation state assessment is greatly reduced, the timeliness of monitoring is improved, and at the same time, the accuracy of transformer monitoring is ensured, thereby greatly improving the efficiency of transformer state assessment.

[0138] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the oil-immersed transformer state assessment method based on graph convolutional neural network in any of the above method embodiments.

[0139] Figure 8 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0140] like Figure 8 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0141] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0142] The communication interface 404 is used for network communication with other devices such as a client or other servers.

[0143] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the oil-immersed transformer state assessment method based on graph convolutional neural network.

[0144] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0145] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0146] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0147] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0148] Acquire time series data of multiple influencing factors associated with preset operating status evaluation parameters of the target transformer within a preset time period, wherein the operating status evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters;

[0149] According to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data;

[0150] Using a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment;

[0151] A plurality of target operating state evaluation parameters are determined according to the prediction data of different influencing factors, and a state evaluation result is determined according to the target operating state evaluation parameters.

[0152] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0153] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network, characterized in that: include: Obtaining time series data of multiple influencing factors associated with preset operating status evaluation parameters of the target transformer within a preset time period, where the operating status evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters; According to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data; Using a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment, wherein the pre-trained influencing factor prediction model is obtained by training a graph convolutional neural network with a training sample set; Determining a plurality of target operating state evaluation parameters according to the prediction data of different influencing factors, and determining a state evaluation result according to the target operating state evaluation parameters; Wherein, the multiple target operation status evaluation parameters are determined based on the prediction data of different influencing factors, including: According to the predicted value of active power on the high voltage side and the predicted value of active power on the low voltage side, the transformer transmission efficiency is calculated, and the transformer transmission efficiency is used as the first target operating state evaluation parameter, wherein the calculation formula is expressed as Wherein, η represents the predicted value of transformer transmission power, P1 represents the predicted value of active power on the high voltage side, and P2 represents the predicted value of active power on the low voltage side; According to the predicted value of oil temperature and the predicted value of ambient temperature, the change rate of oil temperature to ambient temperature is calculated, and the change rate of oil temperature to ambient temperature is used as the second target operating state evaluation parameter, wherein the calculation formula is expressed as follows: Wherein, λ represents the rate of change of oil temperature with respect to ambient temperature, Δθ represents the predicted value of the change of top layer oil temperature during a preset period, and ΔT represents the predicted value of the change of external ambient temperature during a preset period; Perform spectrum analysis on the vibration signal and separate the amplitude of the nth harmonic with a fundamental frequency of 100 Hz; The signal frequency concentration is calculated according to the amplitude of the nth harmonic with the fundamental frequency of 100 Hz, and the signal frequency concentration is used as the third target operating state evaluation parameter; wherein the calculation formula of the signal frequency concentration is expressed as: DFC represents the signal frequency concentration, a 100n Indicates the amplitude of the nth harmonic with a fundamental frequency of 100 Hz, a 100 Indicates the vibration amplitude of the component with a fundamental frequency of 100 Hz.

2. The method according to claim 1, characterized in that According to the correlation between various influencing factors, the data at each time point is used as a graph state, and the time series data is converted into unweighted undirected graph structure time series data, including: Take each influencing factor as a node, determine the edge between nodes according to the correlation between the influencing factors corresponding to each two nodes, and construct an unweighted undirected graph; According to the adjacency matrix of the unweighted undirected graph, the influencing factors corresponding to each time point in the time series data are sequentially introduced into the unweighted undirected graph to obtain the unweighted undirected graph structure time series data.

3. The method according to claim 2, characterized in that Before using the pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data, the method further includes: Acquire historical time series data of influencing factors, take the historical data of the influencing factors in a target period as target input, and take the historical data of the influencing factors at the next moment of the target period as prediction labels to construct training samples, thereby obtaining a training sample set; According to a preset time axis division ratio, the training sample set is divided into a training set and a test set from a time dimension; Using the training set to train the graph convolutional neural network model to obtain an evaluation model; The test set is used to perform model performance testing and parameter optimization on the evaluation model to obtain a pre-trained influencing factor prediction model.

4. The method according to claim 1, characterized in that: The influencing factors also include high-voltage side current and low-voltage side current. Before determining a plurality of target operating state evaluation parameters based on the prediction data of different influencing factors, the method further includes: The high-voltage side active power verification value is calculated based on the high-voltage side current prediction value and the current high-voltage side voltage value; The low-voltage side active power verification value is calculated based on the low-voltage side current prediction value and the current low-voltage side voltage value; The high-voltage side active power prediction value is verified according to the high-voltage side active power verification value, and the high-voltage side active power prediction value is verified according to the low-voltage side active power verification value. If the result of any of the verifications is failed, return to the step of obtaining the time series data of multiple influencing factors associated with the preset operating status evaluation parameters of the target transformer within the preset time period.

5. The method according to any one of claims 1 to 4, characterized in that The status assessment results include normal, caution, abnormal and severe, and the status assessment results are determined according to the target operation status assessment parameters, including: Calculate the difference between each target operating status assessment parameter and the corresponding parameter limit and the fluctuation within a preset time interval; If the difference value of any target operating status evaluation parameter represents an abnormal difference direction, and the absolute value of the difference value is greater than the first difference threshold, then the status evaluation result is determined to be severe; If the fluctuation amount of any target operating state evaluation parameter is greater than the fluctuation threshold, and the absolute value of the difference value is less than the second difference threshold, then determining that the state evaluation result is abnormal; If the difference value of any target operating status evaluation parameter represents a normal difference direction, and the difference value is less than the second difference threshold, determining that the status evaluation result is warning; If the fluctuation amount of any target operating status evaluation parameter is less than or equal to the fluctuation threshold, and the difference value represents a normal difference direction, then the status evaluation result is determined to be normal.

6. A device for evaluating the state of an oil-immersed transformer based on a graph convolutional neural network, characterized in that: include: An acquisition module is used to acquire time series data of multiple influencing factors associated with preset operating status evaluation parameters of the target transformer within a preset time period, where the operating status evaluation parameters include dynamic electrical quantity dimension parameters, oil temperature dimension parameters, and vibration signal dimension parameters; A conversion module, for converting the time series data into time series data with an unweighted undirected graph structure according to the correlation between various influencing factors and taking the data at each time point as a graph state; A processing module, used to use a pre-trained influencing factor prediction model to perform prediction processing on the unweighted undirected graph structure time series data to obtain prediction data of different influencing factors at the next moment, wherein the pre-trained influencing factor prediction model is obtained by training a graph convolutional neural network with a training sample set; A determination module, used to determine a plurality of target operating state evaluation parameters according to the prediction data of different influencing factors, and determine a state evaluation result according to the target operating state evaluation parameters; Wherein, the multiple target operation status evaluation parameters are determined based on the prediction data of different influencing factors, including: According to the predicted value of active power on the high voltage side and the predicted value of active power on the low voltage side, the transformer transmission efficiency is calculated, and the transformer transmission efficiency is used as the first target operating state evaluation parameter, wherein the calculation formula is expressed as Wherein, η represents the predicted value of transformer transmission power, P1 represents the predicted value of active power on the high voltage side, and P2 represents the predicted value of active power on the low voltage side; According to the predicted value of oil temperature and the predicted value of ambient temperature, the change rate of oil temperature to ambient temperature is calculated, and the change rate of oil temperature to ambient temperature is used as the second target operating state evaluation parameter, wherein the calculation formula is expressed as follows: Wherein, λ represents the rate of change of oil temperature with respect to ambient temperature, Δθ represents the predicted value of the change of top layer oil temperature during a preset period, and ΔT represents the predicted value of the change of external ambient temperature during a preset period; Perform spectrum analysis on the vibration signal and separate the amplitude of the nth harmonic with a fundamental frequency of 100 Hz; The signal frequency concentration is calculated according to the amplitude of the nth harmonic with the fundamental frequency of 100 Hz, and the signal frequency concentration is used as the third target operating state evaluation parameter; wherein the calculation formula of the signal frequency concentration is expressed as: DFC represents the signal frequency concentration, a 100n Indicates the amplitude of the nth harmonic with a fundamental frequency of 100 Hz, a 100 Indicates the vibration amplitude of the component with a fundamental frequency of 100 Hz.

7. A storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to perform operations corresponding to the oil-immersed transformer state assessment method based on a graph convolutional neural network as described in any one of claims 1 to 5.

8. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the oil-immersed transformer state assessment method based on graph convolutional neural network as described in any one of claims 1-5.

Images