Method and device for predicting service life of substation equipment

Abstract

The invention provides a method and device for predicting the service life of substation equipment, and the method comprises the steps: constructing a first set containing multi-dimensional features, and iteratively executing the following steps until the first set is empty: placing the feature with the minimum feature weight in the first set in a second set, and setting the deletion probability of the feature with the minimum feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the minimum feature weight from the first set, and updating the feature weight of each feature in the first set; normalizing the deletion probability of each feature in the second set after iteration is completed, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining a target feature based on the selection probability of each feature; and determining the estimated service life of the to-be-predicted equipment based on the feature data of the target feature corresponding to the to-be-predicted equipment, thereby improving the accuracy of the service life prediction task of the substation equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment evaluation, and in particular to a method and device for predicting the service life of substation equipment. Background Art

[0002] As the core hub of the UHV power transmission system, UHV substations undertake the key tasks of transmitting, distributing and regulating electricity. Among them, the stability and safety of equipment such as transformers, reactors and capacitors contained in substations are crucial. Therefore, predicting the remaining service life of substation equipment helps to achieve accurate operation and maintenance of equipment, avoid sudden equipment failures, and reduce maintenance costs. In addition, accurate service life prediction can also optimize asset investment decisions and avoid losses caused by replacing equipment too early or too late.

[0003] The current life prediction methods for substation equipment mainly include: methods based on physical models and methods based on machine learning / deep learning. Among them, the method based on physical models establishes a life prediction model based on the physical characteristics of equipment operation (such as material aging, fatigue curve, thermal effect, etc.), and the method based on machine learning / deep learning uses large-scale monitoring data (such as temperature, humidity, current, voltage, vibration signals, etc.) to train regression or classification models to predict the remaining life of the equipment. Due to the complexity of physical model construction, high requirements on internal parameters and physical characteristics of the equipment, and difficulty in adapting to different equipment and environments, the existing life prediction methods mostly use methods based on machine learning / deep learning. However, there are many features related to the service life of substation equipment, such as environmental data, manufacturer data, maintenance data, and various operating data monitored during actual operation, which may more or less affect the service life of substation equipment. Therefore, the method based on machine learning / deep learning will encounter the challenge of too many input features. The model is easy to learn irrelevant or weakly related information, which increases the computational complexity and reduces the prediction accuracy. Summary of the invention

[0004] The present invention provides a method and device for predicting the service life of substation equipment, which are used to solve the defects of high calculation complexity and low prediction accuracy in the prior art when facing more input features.

[0005] The present invention provides a method for predicting the service life of substation equipment, comprising:

[0006] Data acquisition step: acquiring feature data corresponding to multidimensional features of sample substation equipment and its expected lifespan, and constructing a first set containing multidimensional features;

[0007] The feature weight calculation step is as follows: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the corresponding multidimensional features of the sample substation equipment and its expected lifespan;

[0008] Iteration step: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature;

[0009] Life prediction step: determining the estimated service life of the equipment to be predicted based on the characteristic data corresponding to the target characteristics of the equipment to be predicted.

[0010] According to a method for predicting the service life of substation equipment provided by the present invention, the feature weights of various features are determined based on the following steps:

[0011] Based on the sum of the current adjustment factor and the current value of the lambda parameter of the Lasso regression model, updating the lambda parameter;

[0012] Based on the characteristic data of the multi-dimensional characteristics of the sample substation equipment and its expected lifespan, the characteristic weight of each characteristic is determined using the Lasso regression model;

[0013] If the number of features with feature weight 0 is greater than or equal to 2, then reduce the current adjustment factor;

[0014] If the number of features with feature weight 0 is less than 2, the current adjustment factor is maintained unchanged.

[0015] According to a method for predicting the service life of substation equipment provided by the present invention, the target feature includes a sensor feature and a non-sensor feature; the step of determining the estimated service life of the equipment to be predicted based on feature data corresponding to the target feature of the equipment to be predicted includes:

[0016] Obtaining an embedding vector of feature data corresponding to the target feature of the device to be predicted based on the embedding layer of the life prediction model, and constructing an embedding matrix based on the embedding vector of feature data corresponding to the sensor feature of the device to be predicted; wherein one row of the embedding matrix is ​​an embedding vector of feature data corresponding to any sensor feature of the device to be predicted;

[0017] Encoding and decoding the embedding matrix based on the encoding and decoding layer of the life prediction model to obtain a feature vector of the embedding matrix;

[0018] Based on the fusion layer of the life prediction model, the embedded vector of the feature data corresponding to the non-sensor feature of the device to be predicted and the feature vector of the embedding matrix are fused to obtain a fusion vector of the device to be predicted;

[0019] Based on the prediction layer of the life prediction model, life prediction is performed on the fusion vector of the device to be predicted to obtain an estimated service life of the device to be predicted.

[0020] According to a method for predicting the service life of substation equipment provided by the present invention, the service life prediction model includes multiple encoding and decoding layers; the encoding and decoding layer based on the service life prediction model encodes and decodes the embedded matrix to obtain a feature vector of the embedded matrix, including:

[0021] Based on the sliding time window, each row of the coding matrix output by the coding module of the previous coding and decoding layer is fused to obtain a fused coding matrix;

[0022] Encoding the fused coding matrix based on the coding module of the current coding and decoding layer to obtain the coding matrix output by the coding module of the current coding and decoding layer;

[0023] Decoding a decoding matrix output by a decoding module of a previous coding layer and a coding matrix output by a coding module of a current coding layer based on a decoding module of a current coding layer, to obtain a decoding matrix output by a decoding module of a current coding layer;

[0024] The decoding matrix output by the decoding module of the last encoding and decoding layer is vectorized to obtain the eigenvector of the embedding matrix.

[0025] According to a method for predicting the service life of substation equipment provided by the present invention, each row of the coding matrix output by the coding module of the previous coding and decoding layer is fused based on the sliding time window to obtain a fused coding matrix, including:

[0026] For any row of the coding matrix output by the coding module of the previous coding and decoding layer, the sub-vector in the current sliding time window is converted into a sub-vector with a length of k based on the preset conversion matrix and then spliced ​​to the fusion vector corresponding to the any row; wherein, the initial value of the fusion vector corresponding to the any row is empty, the length of any row of the coding matrix is ​​P×k, the length of the sliding time window is 2k, and the sliding step of the sliding time window is k; the fusion vector corresponding to each row of the coding matrix output by the coding module of the previous coding and decoding layer constitutes the fused coding matrix.

[0027] According to a method for predicting the service life of substation equipment provided by the present invention, the coding module based on the current coding and decoding layer encodes the fused coding matrix to obtain the coding matrix output by the coding module of the current coding and decoding layer, including:

[0028] Based on the first self-attention unit of the encoding module of the current encoding and decoding layer, respectively obtain the first attention vector corresponding to each row of the fused encoding matrix, and construct a first attention matrix based on the first attention vector corresponding to each row of the fused encoding matrix;

[0029] Based on the second self-attention unit of the encoding module of the current codec layer, the second attention vector corresponding to each column of the first attention matrix is ​​obtained respectively, and the encoding matrix is ​​constructed based on the second attention vector corresponding to each column of the first attention matrix.

[0030] According to a method for predicting the service life of substation equipment provided by the present invention, the decoding module based on the current coding and decoding layer decodes the decoding matrix output by the decoding module of the previous coding and decoding layer and the coding matrix output by the coding module of the current coding and decoding layer to obtain the decoding matrix output by the decoding module of the current coding and decoding layer, including:

[0031] Based on the third self-attention unit of the decoding module of the current codec layer, the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer is obtained respectively, and the second attention matrix is ​​constructed based on the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer;

[0032] Based on the fourth self-attention unit of the decoding module of the current codec layer, respectively obtain the fourth attention vector corresponding to each column of the second attention matrix, and construct a basic decoding matrix based on the fourth attention vector corresponding to each column of the second attention matrix;

[0033] Based on the cross attention unit of the decoding module of the current codec layer, an attention transformation is performed on the basic decoding matrix and the encoding matrix output by the encoding module of the current codec layer to obtain the decoding matrix.

[0034] The present invention also provides a device for predicting the service life of substation equipment, comprising:

[0035] A data acquisition unit, used to acquire feature data corresponding to multidimensional features of sample substation equipment and its expected life span, and construct a first set containing multidimensional features;

[0036] A target feature screening unit is used to execute a feature weight calculation step and an iteration step; wherein the feature weight calculation step includes: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the multidimensional features corresponding to the sample substation equipment and its expected life; the iteration step includes: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature;

[0037] The life prediction unit is used to determine the estimated service life of the device to be predicted based on the characteristic data corresponding to the target characteristic of the device to be predicted.

[0038] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for predicting the service life of substation equipment as described above is implemented.

[0039] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for predicting the service life of substation equipment as described in any one of the above is implemented.

[0040] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for predicting the service life of substation equipment.

[0041] The method and device for predicting the service life of substation equipment provided by the present invention obtain feature data and expected life of multidimensional features corresponding to sample substation equipment, and construct a first set containing multidimensional features, and iteratively execute the following steps until the first set is empty: the feature with the smallest feature weight in the first set is placed in the second set and the deletion probability of the feature with the smallest feature weight is set to 0; the deletion probability of each feature in the second set is increased by 1; the feature with the smallest feature weight is deleted from the first set, and the feature weights of each feature in the first set are updated; after the iteration is completed, the deletion probability of each feature in the second set is normalized, and based on the normalized deletion probability of each feature in the second set, the selection probability of each feature is determined; based on the selection probability of each feature, the target feature is determined; thereby, based on the feature data corresponding to the target feature of the equipment to be predicted, the estimated service life of the equipment to be predicted is determined, and the key features that determine the service life of the substation equipment can be screened out, thereby improving the accuracy of the task of predicting the service life of the substation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 It is a flow chart of a method for predicting the service life of substation equipment provided by the present invention;

[0044] Figure 2 It is a structural schematic diagram of the life prediction model provided by the present invention;

[0045] Figure 3 It is a schematic diagram of the flow chart of the encoding and decoding method provided by the present invention;

[0046] Figure 4 It is a structural schematic diagram of a device for predicting the service life of substation equipment provided by the present invention;

[0047] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Figure 1 is a flow chart of a method for predicting the service life of substation equipment provided by the present invention, such as Figure 1 As shown, the method includes:

[0050] Data acquisition step 110: acquiring feature data corresponding to multidimensional features of sample substation equipment and its expected lifespan, and constructing a first set including multidimensional features;

[0051] Feature weight calculation step 120: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the multidimensional features corresponding to the sample substation equipment and its expected lifespan;

[0052] Iteration step 130: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature;

[0053] Life prediction step 140: determining the estimated service life of the device to be predicted based on the characteristic data corresponding to the target characteristic of the device to be predicted.

[0054] Here, feature data corresponding to the multidimensional features of the sample substation equipment are obtained, including environmental data, maintenance data, and operation data monitored during the actual operation of the equipment. Taking the transformer as an example, the multidimensional features may include ambient temperature, ambient humidity, ambient pressure, maintenance object, maintenance time, transformer manufacturing year, voltage, current, load, and oil temperature. It should be noted that the multidimensional features include sensor features, and the feature data of the sensor features (such as ambient temperature, ambient humidity, ambient pressure, voltage, current, load, and oil temperature) are generally time series data, that is, the data sequence monitored by the corresponding sensor within a period of time is recorded. The expected life of the sample substation equipment can be obtained according to the actual retirement life of the corresponding equipment, where the expected life can be a time interval, such as 3.5-4 years. For example, based on the time difference between the actual retirement life of the sample substation equipment and the acquisition time of the feature data obtained, according to the pre-defined time interval, the time interval where the time difference is located is used as the expected life of the equipment. Subsequently, a first set is constructed for the multidimensional features. Among them, the first set can only contain the unique mark of each feature, which is used to indicate the corresponding feature.

[0055] In order to select the features that determine the remaining service life of substation equipment from the multidimensional features, the following feature weight calculation steps can be iteratively performed: the feature with the smallest feature weight in the first set is placed in the second set and the deletion probability of the feature with the smallest feature weight is set to 0; the deletion probability of each feature in the second set is increased by 1; the feature with the smallest feature weight is deleted from the first set, and the feature weights of each feature in the first set are updated. It can be seen that after multiple rounds of iterations, the features with lower feature weights are placed in the second set earlier, and the features placed in the second set earlier have a greater probability of deletion. Among them, when determining the initial feature weights of each feature in the first set and updating the feature weights of each feature in the first set, the feature weight calculation module can be called to input the feature data and expected life of the features of the sample substation equipment corresponding to the features that need to calculate / update the feature weights into the feature weight calculation module to realize the calculation of feature weights.

[0056] In some embodiments, each time the feature weight calculation module is called, the feature weight calculation module may perform the following steps:

[0057] Update the λ parameter based on the sum of the current adjustment factor and the current value of the λ parameter of the Lasso regression model. Subsequently, based on the feature data and expected life of the features of the sample substation equipment corresponding to the features for which the feature weights need to be calculated / updated, which are input into the module, the feature weights of each feature are determined using the Lasso regression model. Among them, features with a feature weight of 0 can be obtained, and the current adjustment factor can be reduced when the number of features with a feature weight of 0 is greater than or equal to 2. If the number of features with a feature weight of 0 is less than 2, the current adjustment factor is maintained unchanged. By setting the current adjustment factor each time the feature weight calculation module is called to control the λ parameter of the Lasso regression model, the sparsity of the Lasso regression model can be flexibly controlled to achieve the effect of compressing only one feature each time to screen out the features with the lowest importance, which helps to improve the accuracy of the probability of deleting each feature in the above-mentioned iterative process, thereby improving the accuracy of feature selection.

[0058] The above-mentioned feature weight calculation step is iterated until the first set is empty, and the iteration is terminated. Subsequently, the deletion probability of each feature in the second set can be normalized, and the selection probability of each feature can be determined based on the normalized deletion probability corresponding to each feature in the second set. Among them, the selection probability of any feature can be obtained by subtracting the normalized deletion probability corresponding to the feature from the value 1. Based on the selection probability of each feature, several features with higher selection probabilities can be determined as target features. In some embodiments, the average value of the selection probability of each feature in the second set can be calculated, and features with a selection probability greater than the average value can be determined as target features.

[0059] After the target features are screened, if a remaining service life prediction is required, the estimated service life of the device to be predicted can be determined based on the feature data of the target features corresponding to the device to be predicted. In some embodiments, the target features include sensor features and non-sensor features, wherein the feature data of the sensor features are time series data, and the non-sensor features are non-time series data. When predicting the life, a linear regression model can be constructed based on the feature weights of each target feature, and the linear regression model can be used to predict the life of the device to be predicted. However, there may be a more complex nonlinear relationship between the various features of the substation equipment, and there may also be a certain correlation in the time series within the sensor features, which may be difficult to reflect by constructing a linear regression model. Therefore, a deep learning mechanism can be used to construct a life prediction model based on the above target features, and the feature data of the sample substation equipment corresponding to each target feature and its expected life can be used to train the life prediction model, so as to use the trained life prediction model to predict the service life. It should be noted that the device to be predicted and the sample substation equipment belong to the same type of equipment.

[0060] In some embodiments, Figure 2 As shown, the life prediction model includes an embedding layer, a codec layer, a fusion layer and a prediction layer. Among them, the embedding layer can be constructed based on the input embedding layer of the Transformer model, and is used to obtain the embedding vector of the feature data of the target feature corresponding to the input device to be predicted. After obtaining the embedding vector of the feature data corresponding to the target feature of the device to be predicted, the embedding layer will also combine the embedding vectors of the feature data of the sensor feature into an embedding matrix, wherein each row of the embedding matrix is ​​an embedding vector of the feature data of a certain sensor feature of the device to be predicted. The codec layer of the life prediction model is used to encode and decode the above-mentioned embedding matrix to obtain the feature vector of the embedding matrix. In one embodiment, the life prediction model may include multiple codec layers connected in sequence, and each codec layer can be constructed based on the Encoder-Decoder model.

[0061] In some embodiments, when there are multiple codec layers, such as Figure 3 As shown, multiple encoding and decoding layers can be encoded and decoded in the following way to obtain the feature vector of the embedding matrix:

[0062] Step 310, fusing each row of the coding matrix output by the coding module of the previous coding / decoding layer based on the sliding time window to obtain a fused coding matrix;

[0063] Step 320, encoding the fused coding matrix based on the coding module of the current coding and decoding layer to obtain the coding matrix output by the coding module of the current coding and decoding layer;

[0064] Step 330: decoding the decoding matrix output by the decoding module of the previous coding layer and the encoding matrix output by the encoding module of the current coding layer based on the decoding module of the current coding layer, to obtain the decoding matrix output by the decoding module of the current coding layer;

[0065] Step 340: vectorize the decoding matrix output by the decoding module of the last encoding and decoding layer to obtain the eigenvector of the embedding matrix.

[0066] Specifically, for any coding layer, each row of the coding matrix output by the coding module of the previous coding layer can be fused based on the sliding time window to obtain a fused coding matrix. Wherein, any row of the coding matrix corresponds to a sensor feature. By fusing each row of the coding matrix row by row, the time information of the same sensor feature on different time scales can be obtained, thereby enhancing the prediction ability of the life prediction model. In some embodiments, for any row of the coding matrix output by the coding module of the previous coding layer, when processing using the sliding time window, the sub-vector in the current sliding time window can be converted into a sub-vector of length k based on a preset conversion matrix (the matrix can be learned during the model training process) and spliced ​​to the fusion vector corresponding to the row to form a new fusion vector. Wherein, the initial value of the fusion vector corresponding to any row is empty, the length of any row of the coding matrix is ​​P×k, the length of the sliding time window is 2k, and the sliding step of the sliding time window is k. Here, if the sliding time window slides to the end of the row, the sliding time window only contains a sub-vector of length k, so the sub-vector can be directly spliced ​​to the fusion vector corresponding to the row to form the final fusion vector of the row. The fusion vector corresponding to each row of the coding matrix output by the coding module of the previous coding and decoding layer constitutes a fusion coding matrix.

[0067] It should be noted that if the current codec layer is the first codec layer, the embedding vector is directly used as the above fusion coding matrix.

[0068] Subsequently, the fused coding matrix is ​​encoded based on the coding module of the current coding layer to obtain the coding matrix output by the coding module of the current coding layer. In some embodiments, considering that it is difficult for the encoder in the existing Encoder-Decoder model to synchronously obtain the relationship between different sensor features, the encoder in the Encoder-Decoder model can be improved to form the coding module in the embodiment of the present application. Specifically, the coding module may include a first self-attention unit and a second self-attention unit.

[0069] Among them, the first self-attention unit is used to obtain the first attention vector corresponding to each row of the fused coding matrix, and construct the first attention matrix based on the first attention vector corresponding to each row of the fused coding matrix. Here, for any row of the fused coding matrix, the first self-attention unit obtains the attention weight of each element of the row based on the self-attention mechanism, normalizes the attention weight of each element of the row, and weights each element of the row to obtain the first attention vector corresponding to the row (i.e., the row vector composed of the weighted elements of the row). Through the first self-attention unit, the temporal relationship within a single sensor feature can be obtained. Thereafter, the second self-attention unit is used to obtain the second attention vector corresponding to each column of the first attention matrix, and construct the coding matrix based on the second attention vector corresponding to each column of the first attention matrix. Here, for any column of the first attention matrix, the second self-attention unit obtains the attention weight of each element of the column based on the self-attention mechanism, normalizes the attention weight of each element of the column, and weights each element of the column to obtain the second attention vector corresponding to the column (i.e., the column vector composed of the weighted elements of the column). Through the second self-attention unit, the correlation between sensor features can be obtained.

[0070] After obtaining the encoding matrix output by the encoding module of the current codec layer, the decoding matrix output by the decoding module of the previous codec layer and the encoding matrix output by the encoding module of the current codec layer can be decoded based on the decoding module of the current codec layer to obtain the decoding matrix output by the encoding module of the current codec layer. It should be noted that if the current codec layer is the first codec layer, the input of the decoding module of the current codec layer is a preset initial matrix (the matrix can be learned through the model training process) and the encoding matrix output by the encoding module of the current codec layer.

[0071] Among them, the decoding module includes a third self-attention unit and a fourth self-attention unit. The third self-attention unit is used to obtain the third attention vector corresponding to each row of the decoding matrix (or the preset initial matrix) output by the decoding module of the previous codec layer, and construct the second attention matrix based on the third attention vector corresponding to each row of the decoding matrix (or the preset initial matrix) output by the decoding module of the previous codec layer. Here, for any row of the decoding matrix (or the preset initial matrix), the third self-attention unit obtains the attention weight of each element of the row based on the self-attention mechanism, normalizes the attention weight of each element of the row, and weights each element of the row to obtain the third attention vector corresponding to the row (that is, the row vector composed of the weighted elements of the row). The fourth self-attention unit is used to obtain the fourth attention vector corresponding to each column of the second attention matrix, and construct the basic decoding matrix based on the fourth attention vector corresponding to each column of the second attention matrix. Here, for any column of the second attention matrix, the fourth self-attention unit obtains the attention weight of each element in the column based on the self-attention mechanism, normalizes the attention weight of each element in the column, and weights each element in the column to obtain the fourth attention vector corresponding to the column (i.e., the column vector composed of the weighted elements in the column).

[0072] The obtained basic decoding matrix contains the correlation between sensor features and the temporal relationship within the sensor features, so the decoding module of the current codec layer can output the basic decoding matrix as a decoding matrix. In order to further improve the accuracy of feature extraction, a cross attention unit can also be added to the decoding module. The cross attention unit is used to perform attention transformation on the basic decoding matrix and the encoding matrix output by the encoding module of the current codec layer based on the cross attention mechanism to obtain a decoding matrix. Among them, a query vector can be constructed based on any row of the basic decoding matrix (assuming it is the i-th row), and a key vector and a value vector can be constructed based on the corresponding row of the encoding matrix output by the encoding module of the current codec layer (also the i-th row), so as to use the attention mechanism to obtain the attention weights of each element in the i-th row of the basic decoding matrix, and weight the elements of the row based on the attention weights of each element in the i-th row of the basic decoding matrix to obtain the fifth attention vector of the i-th row of the basic decoding matrix. The fifth attention vectors of each row of the basic decoding matrix can be combined into a decoding matrix.

[0073] After obtaining the decoding matrix output by the decoding matrix of each coding and decoding layer in the above manner, the decoding matrix output by the decoding module of the last coding and decoding layer can be vectorized to obtain the feature vector of the embedding matrix.

[0074] According to the eigenvector of the embedding matrix, the fusion layer of the life prediction model can be used to fuse the embedding vector of the feature data corresponding to the non-sensor feature of the device to be predicted and the eigenvector of the embedding matrix to obtain the fusion vector of the device to be predicted. Here, the fusion layer can achieve the above fusion by splicing. Subsequently, the prediction layer of the life prediction model can perform life prediction on the fusion vector of the device to be predicted to obtain the estimated service life of the device to be predicted. Among them, the prediction layer can be constructed based on the fully connected layer and the Softmax layer, and the prediction layer can output the probability of the device to be predicted corresponding to each pre-defined time interval, so as to determine the time interval with the highest probability as the estimated service life of the device to be predicted.

[0075] In summary, the method provided by the embodiment of the present invention obtains the feature data and its expected life of the multidimensional features corresponding to the sample substation equipment, and constructs a first set containing multidimensional features, and iteratively executes the following steps until the first set is empty: the feature with the smallest feature weight in the first set is placed in the second set and the deletion probability of the feature with the smallest feature weight is set to 0; the deletion probability of each feature in the second set is increased by 1; the feature with the smallest feature weight is deleted from the first set, and the feature weights of each feature in the first set are updated; after the iteration is completed, the deletion probability of each feature in the second set is normalized, and based on the normalized deletion probability of each feature in the second set, the selection probability of each feature is determined; based on the selection probability of each feature, the target feature is determined; thereby, based on the feature data corresponding to the target feature of the equipment to be predicted, the estimated service life of the equipment to be predicted is determined, the key features that determine the service life of the substation equipment can be screened out, and the accuracy of the substation equipment service life prediction task is improved.

[0076] The following is a description of the substation equipment service life prediction device provided by the present invention. The substation equipment service life prediction device described below and the substation equipment service life prediction method described above can be referenced to each other.

[0077] Based on any of the above embodiments, Figure 4 Schematic diagram of the structure of the device for predicting the service life of substation equipment provided by the present invention. Figure 4 As shown, the device comprises:

[0078] The data acquisition unit 410 is used to acquire feature data corresponding to multi-dimensional features of sample substation equipment and its expected life span, and construct a first set including multi-dimensional features;

[0079] The target feature screening unit 420 is used to perform a feature weight calculation step and an iteration step; wherein the feature weight calculation step includes: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the multidimensional features corresponding to the sample substation equipment and its expected life; the iteration step includes: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature;

[0080] The life prediction unit 430 is used to determine the estimated service life of the device to be predicted based on the feature data corresponding to the target feature of the device to be predicted.

[0081] The device provided by the embodiment of the present invention obtains feature data and its expected life of multidimensional features corresponding to sample substation equipment, and constructs a first set containing multidimensional features, and iteratively executes the following steps until the first set is empty: the feature with the smallest feature weight in the first set is placed in the second set and the deletion probability of the feature with the smallest feature weight is set to 0; the deletion probability of each feature in the second set is increased by 1; the feature with the smallest feature weight is deleted from the first set, and the feature weights of each feature in the first set are updated; after the iteration is completed, the deletion probability of each feature in the second set is normalized, and based on the normalized deletion probability of each feature in the second set, the selection probability of each feature is determined; based on the selection probability of each feature, the target feature is determined; thereby, based on the feature data corresponding to the target feature of the equipment to be predicted, the estimated service life of the equipment to be predicted is determined, the key features that determine the service life of the substation equipment can be screened out, and the accuracy of the substation equipment service life prediction task is improved.

[0082] Based on any of the above embodiments, the feature weight of each feature is determined based on the following steps:

[0083] Based on the sum of the current adjustment factor and the current value of the lambda parameter of the Lasso regression model, updating the lambda parameter;

[0084] Based on the characteristic data of the multi-dimensional characteristics of the sample substation equipment and its expected lifespan, the characteristic weight of each characteristic is determined using the Lasso regression model;

[0085] If the number of features with feature weight 0 is greater than or equal to 2, then reduce the current adjustment factor;

[0086] If the number of features with feature weight 0 is less than 2, the current adjustment factor is maintained unchanged.

[0087] Based on any of the above embodiments, the target feature includes a sensor feature and a non-sensor feature; and determining the estimated service life of the device to be predicted based on the feature data corresponding to the target feature of the device to be predicted includes:

[0088] Obtaining an embedding vector of feature data corresponding to the target feature of the device to be predicted based on the embedding layer of the life prediction model, and constructing an embedding matrix based on the embedding vector of feature data corresponding to the sensor feature of the device to be predicted; wherein one row of the embedding matrix is ​​an embedding vector of feature data corresponding to any sensor feature of the device to be predicted;

[0089] Encoding and decoding the embedding matrix based on the encoding and decoding layer of the life prediction model to obtain a feature vector of the embedding matrix;

[0090] Based on the fusion layer of the life prediction model, the embedded vector of the feature data corresponding to the non-sensor feature of the device to be predicted and the feature vector of the embedding matrix are fused to obtain a fusion vector of the device to be predicted;

[0091] Based on the prediction layer of the life prediction model, life prediction is performed on the fusion vector of the device to be predicted to obtain an estimated service life of the device to be predicted.

[0092] Based on any of the above embodiments, the life prediction model includes multiple encoding and decoding layers; the encoding and decoding layer based on the life prediction model encodes and decodes the embedding matrix to obtain a feature vector of the embedding matrix, including:

[0093] Based on the sliding time window, each row of the coding matrix output by the coding module of the previous coding and decoding layer is fused to obtain a fused coding matrix;

[0094] Encoding the fused coding matrix based on the coding module of the current coding and decoding layer to obtain the coding matrix output by the coding module of the current coding and decoding layer;

[0095] Decoding a decoding matrix output by a decoding module of a previous coding layer and a coding matrix output by a coding module of a current coding layer based on a decoding module of a current coding layer, to obtain a decoding matrix output by a decoding module of a current coding layer;

[0096] The decoding matrix output by the decoding module of the last encoding and decoding layer is vectorized to obtain the eigenvector of the embedding matrix.

[0097] Based on any of the foregoing embodiments, the step of fusing each row of the coding matrix output by the coding module of the previous coding / decoding layer based on the sliding time window to obtain a fused coding matrix includes:

[0098] For any row of the coding matrix output by the coding module of the previous coding and decoding layer, the sub-vector in the current sliding time window is converted into a sub-vector with a length of k based on the preset conversion matrix and then spliced ​​to the fusion vector corresponding to the any row; wherein, the initial value of the fusion vector corresponding to the any row is empty, the length of any row of the coding matrix is ​​P×k, the length of the sliding time window is 2k, and the sliding step of the sliding time window is k; the fusion vector corresponding to each row of the coding matrix output by the coding module of the previous coding and decoding layer constitutes the fused coding matrix.

[0099] Based on any of the foregoing embodiments, the encoding module based on the current encoding and decoding layer encodes the fused encoding matrix to obtain the encoding matrix output by the encoding module of the current encoding and decoding layer, including:

[0100] Based on the first self-attention unit of the encoding module of the current encoding and decoding layer, respectively obtain the first attention vector corresponding to each row of the fused encoding matrix, and construct a first attention matrix based on the first attention vector corresponding to each row of the fused encoding matrix;

[0101] Based on the second self-attention unit of the encoding module of the current codec layer, the second attention vector corresponding to each column of the first attention matrix is ​​obtained respectively, and the encoding matrix is ​​constructed based on the second attention vector corresponding to each column of the first attention matrix.

[0102] Based on any of the foregoing embodiments, the decoding module based on the current coding and decoding layer decodes the decoding matrix output by the decoding module of the previous coding and decoding layer and the encoding matrix output by the encoding module of the current coding and decoding layer to obtain the decoding matrix output by the decoding module of the current coding and decoding layer, including:

[0103] Based on the third self-attention unit of the decoding module of the current codec layer, the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer is obtained respectively, and the second attention matrix is ​​constructed based on the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer;

[0104] Based on the fourth self-attention unit of the decoding module of the current codec layer, respectively obtain the fourth attention vector corresponding to each column of the second attention matrix, and construct a basic decoding matrix based on the fourth attention vector corresponding to each column of the second attention matrix;

[0105] Based on the cross attention unit of the decoding module of the current codec layer, an attention transformation is performed on the basic decoding matrix and the encoding matrix output by the encoding module of the current codec layer to obtain the decoding matrix.

[0106] Figure 5 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 5 As shown, the electronic device may include: a processor (processor) 510, a memory (memory) 520, a communication interface (Communications Interface) 530 and a communication bus 540, wherein the processor 510, the memory 520, and the communication interface 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 520 to execute the substation equipment service life prediction method, which includes: a data acquisition step: obtaining feature data and its expected life of the multidimensional features corresponding to the sample substation equipment, and constructing a first set containing the multidimensional features; a feature weight calculation step: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data and its expected life of the multidimensional features corresponding to the sample substation equipment; an iteration step: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature; a life prediction step: determining the estimated service life of the device to be predicted based on the feature data of the target feature corresponding to the device to be predicted.

[0107] In addition, the logic instructions in the above-mentioned memory 520 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0108] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the substation equipment service life prediction method provided by the above methods, and the method includes: a data acquisition step: obtaining feature data and its expected life corresponding to multidimensional features of sample substation equipment, and constructing a first set containing multidimensional features; a feature weight calculation step: placing the feature with the smallest feature weight in the first set into a second set and setting the deletion probability of the feature with the smallest feature weight to 0; setting the deletion probability of each feature in the second set to zero; The rate is increased by 1; the feature with the smallest feature weight is deleted from the first set, and the feature weight of each feature in the first set is updated; wherein the feature weight of each feature is determined based on the feature data of the multidimensional features corresponding to the sample substation equipment and its expected life; iterative step: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature; life prediction step: determining the estimated service life of the equipment to be predicted based on the feature data of the target feature corresponding to the equipment to be predicted.

[0109] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented by a processor to execute the above-mentioned substation equipment service life prediction method, the method comprising: a data acquisition step: acquiring feature data and its expected life of the corresponding multidimensional features of the sample substation equipment, and constructing a first set containing the multidimensional features; a feature weight calculation step: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data and its expected life of the corresponding multidimensional features of the sample substation equipment; an iteration step: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature; a life prediction step: determining the estimated service life of the device to be predicted based on the feature data of the target feature corresponding to the device to be predicted.

[0110] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0111] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the service life of substation equipment, characterized in that: include: Data acquisition step: acquiring feature data corresponding to multidimensional features of sample substation equipment and its expected lifespan, and constructing a first set containing multidimensional features; The feature weight calculation step is as follows: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the corresponding multidimensional features of the sample substation equipment and its expected lifespan; Iteration step: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature; Life prediction step: determining the estimated service life of the equipment to be predicted based on the characteristic data corresponding to the target characteristics of the equipment to be predicted.

2. The method for predicting the service life of substation equipment according to claim 1, characterized in that: The feature weights of each feature are determined based on the following steps: Based on the sum of the current adjustment factor and the current value of the lambda parameter of the Lasso regression model, updating the lambda parameter; Based on the characteristic data of the multi-dimensional characteristics of the sample substation equipment and its expected lifespan, the characteristic weight of each characteristic is determined using the Lasso regression model; If the number of features with feature weight 0 is greater than or equal to 2, then reduce the current adjustment factor; If the number of features with feature weight 0 is less than 2, the current adjustment factor is maintained unchanged.

3. The method for predicting the service life of substation equipment according to claim 1 or 2, characterized in that: The target features include sensor features and non-sensor features; The determining the estimated service life of the equipment to be predicted based on the feature data corresponding to the target feature of the equipment to be predicted includes: Obtaining an embedding vector of feature data corresponding to the target feature of the device to be predicted based on the embedding layer of the life prediction model, and constructing an embedding matrix based on the embedding vector of feature data corresponding to the sensor feature of the device to be predicted; wherein one row of the embedding matrix is ​​an embedding vector of feature data corresponding to any sensor feature of the device to be predicted; Encoding and decoding the embedding matrix based on the encoding and decoding layer of the life prediction model to obtain a feature vector of the embedding matrix; Based on the fusion layer of the life prediction model, the embedded vector of the feature data corresponding to the non-sensor feature of the device to be predicted and the feature vector of the embedding matrix are fused to obtain a fusion vector of the device to be predicted; Based on the prediction layer of the life prediction model, life prediction is performed on the fusion vector of the device to be predicted to obtain an estimated service life of the device to be predicted.

4. The method for predicting the service life of substation equipment according to claim 3, characterized in that: The life prediction model comprises a plurality of encoding and decoding layers; the encoding and decoding layer based on the life prediction model encodes and decodes the embedding matrix to obtain a feature vector of the embedding matrix, including: Based on the sliding time window, each row of the coding matrix output by the coding module of the previous coding and decoding layer is fused to obtain a fused coding matrix; Encoding the fused coding matrix based on the coding module of the current coding and decoding layer to obtain the coding matrix output by the coding module of the current coding and decoding layer; Decoding a decoding matrix output by a decoding module of a previous coding layer and a coding matrix output by a coding module of a current coding layer based on a decoding module of a current coding layer, to obtain a decoding matrix output by a decoding module of a current coding layer; The decoding matrix output by the decoding module of the last encoding and decoding layer is vectorized to obtain the eigenvector of the embedding matrix.

5. The method for predicting the service life of substation equipment according to claim 4, characterized in that: The step of fusing each row of the coding matrix output by the coding module of the previous coding / decoding layer based on the sliding time window to obtain a fused coding matrix includes: For any row of the coding matrix output by the coding module of the previous coding and decoding layer, the sub-vector in the current sliding time window is converted into a sub-vector with a length of k based on the preset conversion matrix and then spliced ​​to the fusion vector corresponding to the any row; wherein, the initial value of the fusion vector corresponding to the any row is empty, the length of any row of the coding matrix is ​​P×k, the length of the sliding time window is 2k, and the sliding step of the sliding time window is k; the fusion vector corresponding to each row of the coding matrix output by the coding module of the previous coding and decoding layer constitutes the fused coding matrix.

6. The method for predicting the service life of substation equipment according to claim 4, characterized in that: The encoding module based on the current encoding and decoding layer encodes the fused encoding matrix to obtain the encoding matrix output by the encoding module of the current encoding and decoding layer, including: Based on the first self-attention unit of the encoding module of the current encoding and decoding layer, respectively obtain the first attention vector corresponding to each row of the fused encoding matrix, and construct a first attention matrix based on the first attention vector corresponding to each row of the fused encoding matrix; Based on the second self-attention unit of the encoding module of the current codec layer, the second attention vector corresponding to each column of the first attention matrix is ​​obtained respectively, and the encoding matrix is ​​constructed based on the second attention vector corresponding to each column of the first attention matrix.

7. The method for predicting the service life of substation equipment according to claim 4, characterized in that: The decoding module based on the current codec layer decodes the decoding matrix output by the decoding module of the previous codec layer and the encoding matrix output by the encoding module of the current codec layer to obtain the decoding matrix output by the decoding module of the current codec layer, including: Based on the third self-attention unit of the decoding module of the current codec layer, the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer is obtained respectively, and the second attention matrix is ​​constructed based on the third attention vector corresponding to each row of the decoding matrix output by the decoding module of the previous codec layer; Based on the fourth self-attention unit of the decoding module of the current codec layer, respectively obtain the fourth attention vector corresponding to each column of the second attention matrix, and construct a basic decoding matrix based on the fourth attention vector corresponding to each column of the second attention matrix; Based on the cross attention unit of the decoding module of the current codec layer, an attention transformation is performed on the basic decoding matrix and the encoding matrix output by the encoding module of the current codec layer to obtain the decoding matrix.

8. A device for predicting the service life of substation equipment, characterized in that: include: A data acquisition unit, used to acquire feature data corresponding to multidimensional features of sample substation equipment and its expected life span, and construct a first set containing multidimensional features; A target feature screening unit is used to execute a feature weight calculation step and an iteration step; wherein the feature weight calculation step includes: placing the feature with the smallest feature weight in the first set into the second set and setting the deletion probability of the feature with the smallest feature weight to 0; adding 1 to the deletion probability of each feature in the second set; deleting the feature with the smallest feature weight from the first set, and updating the feature weights of each feature in the first set; wherein the feature weight of each feature is determined based on the feature data of the multidimensional features corresponding to the sample substation equipment and its expected life; the iteration step includes: repeating the feature weight calculation step until the first set is empty; normalizing the deletion probability of each feature in the second set, and determining the selection probability of each feature based on the normalized deletion probability of each feature in the second set; determining the target feature based on the selection probability of each feature; The life prediction unit is used to determine the estimated service life of the device to be predicted based on the characteristic data corresponding to the target characteristic of the device to be predicted.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for predicting the service life of substation equipment according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the service life of substation equipment according to any one of claims 1 to 7 is implemented.

Images