Training of amplitude processing network and amplitude processing method, device and storage medium thereof

Abstract

The application discloses a kind of amplitude processing network training and its amplitude processing method, equipment and storage medium, the method comprises: when collecting original amplitude data to pumped storage unit on multiple time, load amplitude processing network;The original amplitude data of each time is respectively input into first encoder and encoded first original amplitude feature;The original amplitude data of multiple times is simultaneously input into second encoder and encoded second original amplitude feature;Multiple first original amplitude features and second original amplitude features are input into feature interaction module and interacted into fusion amplitude feature;Fusion amplitude feature is input into decoder and decoded into target amplitude feature;Target amplitude feature is input into multiple task processing module and generates multiple amplitude task information;According to multiple amplitude task information, generate total loss value;According to total loss value, update amplitude processing network.This embodiment improves the robustness of amplitude processing network, improves the accuracy of multi-task processing.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, and in particular to a training method, device and storage medium for amplitude processing networks. Background Technology

[0002] A pumped storage power plant is a complex automated intelligent operating system. The system has multiple core devices such as a water turbine pumped storage unit and a pumped storage unit. The pumped storage unit has a variety of important equipment such as an upper frame, a lower frame, a water guide top cover, and a thrust bearing. During normal operation, these devices will generate vibration amplitudes with certain changing patterns.

[0003] During the operation of pumped storage power plants, different complex operating conditions occur at different times. To ensure the stable and safe operation of pumped storage units, threshold methods and Seq2Seq (sequence to sequence) are currently used to monitor the amplitude of equipment such as the upper frame, lower frame, water guide top cover, and thrust bearing. Corresponding safety measures are formulated and quickly activated to ensure the safe operation of pumped storage units and avoid the occurrence of sudden accidents.

[0004] The threshold method involves triggering an alarm or fault signal when the vibration amplitude exceeds a certain threshold. However, the threshold is fixed. If the threshold range is too narrow, false alarms are likely to occur; if the threshold range is too wide, the sensitivity is low, resulting in low efficiency of the threshold method.

[0005] Seq2Seq is typically a time-series model based on RNN (Recurrent Neural Network), which connects the encoder and decoder serially. This reduces the amount of feature information extracted, resulting in lower robustness of the time-series model. Summary of the Invention

[0006] In view of this, the present invention provides a training method, device and storage medium for amplitude processing network, and an amplitude processing method thereof, in order to improve the efficiency and robustness of monitoring the amplitude of pumped storage units.

[0007] A first aspect of the present invention provides a method for training an amplitude processing network, comprising:

[0008] When raw amplitude data of the pumped storage unit is collected at multiple times, an amplitude processing network is loaded; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0009] The original amplitude data at each time point are respectively input into the first encoder to encode the first original amplitude feature;

[0010] The original amplitude data at multiple times are simultaneously input into the second encoder to encode the second original amplitude feature;

[0011] Multiple first original amplitude features and second original amplitude features are input into the feature interaction module to interact and form a fused amplitude feature;

[0012] The fused amplitude features are input into the decoder and decoded into target amplitude features;

[0013] The target amplitude feature is input into the multi-task processing module to generate multiple amplitude task information;

[0014] A total loss value is generated based on multiple amplitude task information.

[0015] The amplitude processing network is updated based on the total loss value.

[0016] A second aspect of the present invention provides an amplitude processing method, comprising:

[0017] When raw amplitude data is collected from the pumped storage unit at multiple times, an amplitude processing network trained as described in the first aspect above is loaded; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0018] The original amplitude data at each time point are respectively input into the first encoder to encode the first original amplitude feature;

[0019] The original amplitude data at multiple times are simultaneously input into the second encoder to encode the second original amplitude feature;

[0020] Multiple first original amplitude features and second original amplitude features are input into the feature interaction module to interact and form a fused amplitude feature;

[0021] The fused amplitude features are input into the decoder and decoded into target amplitude features;

[0022] The target amplitude feature is input into the multi-task processing module to generate multiple amplitude task information.

[0023] A third aspect of the present invention provides a training apparatus for an amplitude processing network, comprising:

[0024] An amplitude processing network loading module is used to load an amplitude processing network when raw amplitude data of a pumped storage unit is collected at multiple times; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0025] The first original amplitude feature encoding module is used to input the original amplitude data at each time moment into the first encoder to encode the first original amplitude feature;

[0026] The second original amplitude feature encoding module is used to simultaneously input the original amplitude data at multiple times into the second encoder to encode the second original amplitude feature;

[0027] The fusion amplitude feature interaction module is used to input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and generate fusion amplitude features;

[0028] The target amplitude feature decoding module is used to input the fused amplitude feature into the decoder and decode it into the target amplitude feature;

[0029] An amplitude task information generation module is used to input the target amplitude features into the multi-task processing module to generate multiple amplitude task information.

[0030] The total loss value generation module is used to generate a total loss value based on multiple amplitude task information.

[0031] An amplitude processing network update module is used to update the amplitude processing network based on the total loss value.

[0032] A fourth aspect of the present invention provides an amplitude processing apparatus, comprising:

[0033] An amplitude processing network loading module is used to load an amplitude processing network trained as described in the first aspect above when raw amplitude data of the pumped storage unit is collected at multiple times; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0034] The first original amplitude feature encoding module is used to input the original amplitude data at each time moment into the first encoder to encode the first original amplitude feature;

[0035] The second original amplitude feature encoding module is used to simultaneously input the original amplitude data at multiple times into the second encoder to encode the second original amplitude feature;

[0036] The fusion amplitude feature interaction module is used to input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and generate fusion amplitude features;

[0037] The target amplitude feature decoding module is used to input the fused amplitude feature into the decoder and decode it into the target amplitude feature;

[0038] An amplitude task information generation module is used to input the target amplitude features into the multi-task processing module to generate multiple amplitude task information.

[0039] A fifth aspect of the present invention provides an electronic device comprising:

[0040] At least one processor; and

[0041] A memory communicatively connected to the at least one processor; wherein,

[0042] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform a training method for an amplitude processing network as described in the first aspect above or an amplitude processing method as described in the second aspect above.

[0043] A sixth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a training method for an amplitude processing network as described in the first aspect above or an amplitude processing method as described in the second aspect above.

[0044] A seventh aspect of the present invention provides a computer program product comprising a computer program that, when executed by a processor, implements a training method for an amplitude processing network as described in the first aspect above or an amplitude processing method as described in the second aspect above.

[0045] In this embodiment, when raw amplitude data is collected from the pumped storage unit at multiple time points, an amplitude processing network is loaded. The amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module. The raw amplitude data at each time point are input into the first encoder to encode first raw amplitude features. The raw amplitude data from multiple time points are simultaneously input into the second encoder to encode second raw amplitude features. The multiple first and second raw amplitude features are input into the feature interaction module to interact and generate a fused amplitude feature. The fused amplitude feature is input into the decoder to decode a target amplitude feature. The target amplitude feature is input into the multi-task processing module to generate multiple amplitude task information. A total loss value is generated based on the multiple amplitude task information. The amplitude processing network is updated based on the total loss value. This embodiment performs feature interaction and fusion between the encoder and decoder, which can increase the information content of the features, thereby improving the robustness of the amplitude processing network and the accuracy of multi-task processing. Furthermore, the adaptive use of amplitude features to perform multi-task processing provides high flexibility, thereby improving processing efficiency.

[0046] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a training method for an amplitude processing network provided in Embodiment 1 of the present invention.

[0049] Figure 2 This is a schematic diagram of the structure of an amplitude processing network provided in Embodiment 1 of the present invention.

[0050] Figure 3 This is a flowchart of an amplitude processing method provided in Embodiment 2 of the present invention.

[0051] Figure 4 This is a schematic diagram of the structure of a training device for an amplitude processing network provided in Embodiment 3 of the present invention.

[0052] Figure 5 This is a schematic diagram of an amplitude processing device provided in Embodiment 4 of the present invention.

[0053] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Embodiment 5 of the present invention. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0055] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate so that the embodiments of the invention described herein can cover implementations in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0056] Example 1

[0057] See Figure 1 The diagram illustrates a flowchart of a training method for an amplitude processing network according to Embodiment 1 of the present invention. This method can be executed by a training device for the amplitude processing network. The training of the amplitude processing network and the amplitude processing device can be implemented in hardware and / or software. The training device for the amplitude processing network can be configured in an electronic device. Figure 1 As shown, the method includes:

[0058] Step 101: When raw amplitude data of the pumped storage unit is collected at multiple times, load the amplitude processing network.

[0059] In this embodiment, vibration sensors can be installed on various equipment such as the upper frame, lower frame, water guide top cover, and thrust bearing of the pumped storage unit. During the operation of the pumped storage unit, the vibration amplitude collected by the vibration sensors at various times can be read and recorded as the original amplitude data.

[0060] like Figure 2 As shown, training data for each batch can be constructed using raw vibration data from multiple time points [{x t-T+1 ,x t-T+2 ,…,x t},x t+1 ,x” t+1 ], where {x t-T+1 ,x t-T+2 ,…,x t} represents the original amplitude data from time t-T+1 to time t, x t+1 The original amplitude data at time t+1, x” t+1 For x t+1 Constructed abnormal amplitude data.

[0061] In this embodiment, an amplitude processing network can be constructed based on deep learning. The amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0062] When training the amplitude processing network, the amplitude processing network is loaded into memory and run.

[0063] Step 102: Input the original amplitude data at each time point into the first encoder to encode the first original amplitude feature.

[0064] In this embodiment, as Figure 2 As shown, the original amplitude data at each time point can be input into the first encoder, and the first encoder encodes the original amplitude data at each time point to obtain the first original amplitude feature C1.

[0065] For example, the first encoder is a backbone network with a self-attention mechanism. Optionally, the backbone network with a self-attention mechanism includes a neural network Transformer, and the backbone network is constructed using Transformer as a module.

[0066] Step 103: Simultaneously input the raw amplitude data from multiple moments into the second encoder to encode the second raw amplitude feature.

[0067] In this embodiment, as Figure 2 As shown, the original amplitude data at multiple times can be input into the second encoder, and the second encoder encodes the original amplitude data at multiple times to obtain the second original amplitude feature C2.

[0068] For example, the second encoder is a backbone network of a recurrent neural network (RNN), and optionally, the backbone network of the recurrent neural network includes a long short-term memory network (LSTM).

[0069] Step 104: Input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and form a fused amplitude feature.

[0070] In this embodiment, as Figure 2 As shown, multiple first original amplitude features C1 and second original amplitude features C2 can be input into the feature interaction module. The feature interaction module will interactively fuse the multiple first original amplitude features C1 and second original amplitude features C2 to obtain the fused amplitude feature C3.

[0071] In one embodiment of the present invention, step 104 may include the following steps:

[0072] Step 1041: Perform a normalization operation on each of the first original amplitude features to obtain the first candidate amplitude features.

[0073] In this embodiment, as Figure 2 As shown, a normalization operation can be performed on each of the first original amplitude features C1 to obtain the first candidate amplitude feature S1.

[0074] For example, such as Figure 2 As shown, if the normalization operation is Instance Normalization, then the first candidate amplitude feature is represented as:

[0075]

[0076] Where i is time, i = 1, 2, ..., T, C1 is the first original amplitude feature, U1 is the average value of the first original amplitude feature, σ1 is the variance of the first original amplitude feature, ε is the hyperparameter (e.g., 0.00001), and S1 is the first candidate amplitude feature.

[0077] Step 1042: Perform a normalization operation on the second original amplitude feature to obtain the second candidate amplitude feature.

[0078] In this embodiment, as Figure 2 As shown, a normalization operation can be performed on each of the second original amplitude features C2 to obtain the second candidate amplitude features S2.

[0079] Generally, the normalization operation performed on each of the first original amplitude features is the same as the normalization operation performed on the second original amplitude features.

[0080] For example, if the normalization operation is Instance Normalization, then the second candidate amplitude feature is represented as:

[0081]

[0082] Where C2 is the second original amplitude feature, U2 is the average value of the second original amplitude feature, σ2 is the variance of the second original amplitude feature, ε is the hyperparameter (e.g., 0.00001), and S2 is the second candidate amplitude feature.

[0083] Step 1043: Calculate the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature.

[0084] In this embodiment, for the same normalization operation, an attention mechanism can be used to calculate the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature.

[0085] In specific implementations, such as Figure 2 As shown, in the attention mechanism, the first candidate amplitude feature S1 is used as the key value vector K and the second candidate amplitude feature S2 is used as the query vector Q, and multiple benchmark weights w are calculated.

[0086] At this point, the key value vector K (i.e., the first candidate amplitude feature S1) and the query vector Q (i.e., the second candidate amplitude feature S2) are multiplied by a dot product. The result of the dot product is normalized by the softmax function to obtain the weights w of each of the multiple reference values.

[0087] Therefore, the multiple benchmark weights w are represented as:

[0088]

[0089] Where K∈R T*D , Q∈R 1*D w∈R T*1 T represents the time period, and D represents the feature dimension of each time series.

[0090] Alignment is performed on the average value. At this point, the product of the average value of the first original amplitude feature and the reference weight is added to obtain the aligned average value. Therefore, the aligned average value u can be expressed as:

[0091] The average value of the second original amplitude feature is linearly fused with the aligned average value to form the feature interaction average value. For example, the feature interaction average value U can be expressed as U = (u + U2) * 0.5.

[0092] When aligning the variances, the products of the variances of the first original amplitude feature and the baseline weights are added together to obtain the aligned variance. The aligned variance σ can then be expressed as:

[0093] The variance of the second original amplitude feature is linearly fused with the aligned variance to form the feature interaction variance. For example, the feature interaction variance U can be expressed as Σ=(σ+σ2)*0.5.

[0094] Step 1044: Align the second candidate vibration features into fused amplitude features based on attention weights.

[0095] In this embodiment, as Figure 2 As shown, the second candidate vibration feature S2 is aligned to the target amplitude feature C3 based on the attention weight.

[0096] In specific implementations, such as Figure 2 As shown, the fused amplitude characteristics are represented as follows:

[0097] C3 = S2 × Σ + U;

[0098] Where C3 is the fused amplitude feature, S2 is the second candidate vibration feature, U is the feature interaction mean, and Σ is the feature interaction variance.

[0099] In this embodiment, based on the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature under the same normalization operation, the second candidate vibration feature is aligned into a fused amplitude feature, which can obtain more robust features.

[0100] Step 105: Input the fused amplitude features into the decoder and decode them into target amplitude features.

[0101] In this embodiment, as Figure 2 As shown, the fused amplitude feature C3 can be input into the decoder, which will decode the fused amplitude feature C3 into the target amplitude feature.

[0102] For example, such as Figure 2 As shown, the decoder is a Long Short-Term Memory (LSTM) network.

[0103] Step 106: Input the target amplitude features into the multi-task processing module to generate multiple amplitude task information.

[0104] In this embodiment, the target amplitude features can be input into the multi-task processing module, which generates multiple amplitude task information based on the target amplitude features according to the business requirements.

[0105] In its implementation, the multi-task processing module includes a sequence prediction module, a working condition classification module, and an anomaly discriminator. Correspondingly, the multiple amplitude task information includes target amplitude data, target working condition status, and anomaly discrimination labels.

[0106] Among them, the sequence prediction module is composed of multiple fully connected layers (FC), the working condition classification module is composed of multiple fully connected layers (FC), and the anomaly discriminator is composed of multiple fully connected layers (FC).

[0107] For example, the sequence prediction module consists of three fully connected layers (FC) with feature sizes of 256, 512, and 2, respectively.

[0108] The working condition classification module consists of three fully connected layers (FC) with feature sizes of 256, 512, and k, respectively, where k is the number of target working condition states (i.e., categories).

[0109] The anomaly discriminator consists of three fully connected layers (FC) with feature sizes of 256, 512, and 2, respectively.

[0110] So, if Figure 2 As shown, the target amplitude features are input into the sequence prediction module to generate the target amplitude data x' for the next time step. t+1 .

[0111] Input the target amplitude characteristics into the operating condition classification module to classify the target operating condition status of the pumped storage unit.

[0112] The target amplitude data is input into the anomaly discriminator to generate anomaly discrimination labels; the anomaly discrimination labels indicate whether the target amplitude data is abnormal.

[0113] Furthermore, for the anomaly detector, real raw amplitude data x can be used. t+1 Construct a batch of abnormal amplitude data x” t+1 .

[0114] The methods used to construct the data include: swapping the column positions of the original amplitude data; changing the value of a certain measurement point in the original amplitude data to make it exceed the normal data range; changing the data of multiple column positions in the original amplitude data to make it fluctuate more, and so on.

[0115] x t+1 Abnormal amplitude data x” t+1 With target amplitude data x' t+1 Calculate the difference between each pair of anomalies, and input the difference values ​​into the anomaly detector for training until the difference (x') is reached. t+1 -x t+1 When inputting, the exception label is normal, and all other differences are considered abnormal.

[0116] Step 107: Generate the total loss value based on multiple amplitude task information.

[0117] In this embodiment, multiple amplitude task information can be substituted into a preset loss function for calculation to generate a total loss value.

[0118] In specific implementations, such as Figure 2 As shown, on the one hand, the loss is calculated based on the difference (such as the mean square error) between the target amplitude data and the original amplitude data at the next moment, thereby generating the first sub-loss value L. 预测loss .

[0119] On the other hand, a true / false discriminator is constructed and trained using a generative adversarial network (GAN) training mode. This discriminator is then used to analyze the predicted target amplitude data x' at time t+1. t+1 Compared with the actual raw amplitude data x t+1 Perform a true / false judgment, that is, judge the current data (i.e., the target amplitude data x'). t+1 Or the actual raw amplitude data x t+1 () is real data or fake data.

[0120] The target amplitude data is input into a preset discriminator to generate a true / false label; the true / false label indicates whether the target amplitude data is true or false.

[0121] A second sub-loss value L is generated based on the true / false labels. gan .

[0122] At this point, the second sub-loss value L gan Represented as:

[0123]

[0124] Where G represents the generator network (i.e., the first encoder, the second encoder, the feature interaction module, the decoder, and the sequence prediction module), D represents the discriminator, and P represents the real / fake discriminator. data (x) represents the distribution of the original amplitude data, P z (z) represents the distribution of the target amplitude data, z represents the original amplitude data, z = {x t-T+1 ,x t-T+2 ,…,x t}, where x represents the actual raw amplitude data at the next moment. t+1 .

[0125] On the other hand, a third sub-loss value L is generated based on the differences (such as cross-entropy) between the target operating condition and the sample operating conditions labeled for the pumped storage unit. 工况Discriminator .

[0126] On the other hand, based on the anomaly detection labels, a fourth sub-loss value L is generated using methods such as binary cross-entropy. 异常 .

[0127] The first, second, third, and fourth sub-loss values ​​are merged into a total loss value using linear or nonlinear methods.

[0128] For example, the total loss value can be expressed as:

[0129] L total =r1*L 预测loss +r2* Lgan +r3*L 工况Discriminator +r4*L 异常

[0130] Among them, L total The total loss value is represented by r1, r2, r3, and r4, which are all hyperparameters.

[0131] Step 108: Update the amplitude processing network based on the total loss value.

[0132] In this embodiment, the total loss value can be substituted into optimization algorithms such as SGD (stochastic gradient descent) and Adam (adaptive momentum) to calculate the update magnitude of the parameters in the amplitude processing network, and the parameters in the amplitude processing network are updated according to the update magnitude.

[0133] In addition, training conditions can be preset as conditions for stopping the training of the amplitude processing network. For example, the number of iterations reaches a certain threshold, the loss value is less than a certain threshold, the change in the loss value in multiple training iterations is less than a certain threshold, and so on. In each round of training iteration, it is determined whether the current corresponding parameters meet the training conditions.

[0134] If the training conditions are met, the amplitude processing network can be considered to have completed training.

[0135] If the training conditions are not met, the next round of iterative training can be started, and steps 102-108 can be executed again. This iterative training can be repeated until the amplitude processing network is trained.

[0136] In this embodiment, when raw amplitude data is collected from the pumped storage unit at multiple time points, an amplitude processing network is loaded. The amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module. The raw amplitude data at each time point are input into the first encoder to encode first raw amplitude features. The raw amplitude data from multiple time points are simultaneously input into the second encoder to encode second raw amplitude features. The multiple first and second raw amplitude features are input into the feature interaction module to interact and generate a fused amplitude feature. The fused amplitude feature is input into the decoder to decode a target amplitude feature. The target amplitude feature is input into the multi-task processing module to generate multiple amplitude task information. A total loss value is generated based on the multiple amplitude task information. The amplitude processing network is updated based on the total loss value. This embodiment performs feature interaction and fusion between the encoder and decoder, which can increase the information content of the features, thereby improving the robustness of the amplitude processing network and the accuracy of multi-task processing. Furthermore, the adaptive use of amplitude features to perform multi-task processing provides high flexibility, thereby improving processing efficiency.

[0137] Example 2

[0138] See Figure 3 The diagram illustrates a flowchart of an amplitude processing method according to Embodiment 1 of the present invention. This method can be executed by an amplitude processing device. The training of the amplitude processing network and the amplitude processing device can be implemented in hardware and / or software, and the amplitude processing device can be configured in an electronic device. Figure 3 As shown, the method includes:

[0139] Step 301: When raw amplitude data of the pumped storage unit is collected at multiple times, load the amplitude processing network.

[0140] In practical applications, if the method of Example 1 is used to train the amplitude processing network and to perform verification and testing on the amplitude processing network, the amplitude processing network can be deployed online.

[0141] The amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module.

[0142] Step 302: Input the original amplitude data at each time point into the first encoder to encode the first original amplitude feature.

[0143] For example, the first encoder is a backbone network that applies a self-attention mechanism, which includes a neural network Transformer.

[0144] Step 303: Simultaneously input the raw amplitude data at multiple times into the second encoder to encode the second raw amplitude feature.

[0145] For example, the second encoder is a long short-term memory network.

[0146] Step 304: Input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and form a fused amplitude feature.

[0147] In one embodiment of the present invention, step 304 may include the following steps:

[0148] Step 3041: Perform a normalization operation on each of the first original amplitude features to obtain the first candidate amplitude features.

[0149] For example, the first candidate amplitude feature is represented as:

[0150]

[0151] Where i is time, C1 is the first original amplitude feature, U1 is the average value of the first original amplitude feature, σ1 is the variance of the first original amplitude feature, ε is the hyperparameter, and S1 is the first candidate amplitude feature.

[0152] Step 3042: Perform a normalization operation on the second original amplitude feature to obtain the second candidate amplitude feature.

[0153] For example, the second candidate amplitude feature is represented as:

[0154]

[0155] Where C2 is the second original amplitude feature, U2 is the average value of the second original amplitude feature, σ2 is the variance of the second original amplitude feature, ε is the hyperparameter, and S2 is the second candidate amplitude feature.

[0156] Step 3043: Calculate the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature.

[0157] In the attention mechanism, multiple baseline weights are calculated using the first candidate amplitude feature as the key value vector and the second candidate amplitude feature as the query vector. The product of the average value of the first original amplitude feature and the baseline weights is added to obtain the aligned average value. The average value of the second original amplitude feature and the aligned average value are linearly fused to obtain the feature interaction average value. The product of the variance of the first original amplitude feature and the baseline weights is added to obtain the aligned variance. The variance of the second original amplitude feature and the aligned variance are linearly fused to obtain the feature interaction variance.

[0158] Step 3044: Align the second candidate vibration features into fused amplitude features based on attention weights.

[0159] For example, the fused amplitude characteristics are represented as:

[0160] C3 = S2 × Σ + U;

[0161] Wherein, C3 is the fused amplitude feature, S2 is the second candidate vibration feature, U is the feature interaction average, and Σ is the feature interaction variance.

[0162] Step 305: Input the fused amplitude features into the decoder and decode them into target amplitude features.

[0163] For example, the decoder is a long short-term memory network.

[0164] Step 306: Input the target amplitude features into the multi-task processing module to generate multiple amplitude task information.

[0165] In its implementation, the multi-task processing module includes a sequence prediction module, a working condition classification module, and an anomaly discriminator. Multiple amplitude task information includes target amplitude data, target working condition status, and anomaly discrimination labels.

[0166] For example, the sequence prediction module consists of multiple fully connected layers, the working condition classification module consists of multiple fully connected layers, and the anomaly discriminator consists of multiple fully connected layers.

[0167] On the one hand, the target amplitude characteristics are input into the sequence prediction module to generate the target amplitude data for the next time step.

[0168] On the other hand, the target amplitude characteristics are input into the operating condition classification module to classify the target operating condition status of the pumped storage unit.

[0169] On the other hand, the target amplitude data is input into the anomaly discriminator to generate anomaly discrimination labels; the anomaly discrimination labels indicate whether the target amplitude data is abnormal.

[0170] In this embodiment, since steps 301-306 are basically similar to the application in Embodiment 1, the description is relatively simple. For relevant details, please refer to the description in Embodiment 1. This embodiment of the present invention will not be described in detail here.

[0171] In this embodiment, when raw amplitude data of the pumped storage unit is collected at multiple times, an amplitude processing network is loaded. The amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder, and a multi-task processing module. The raw amplitude data at each time point are input into the first encoder to encode first raw amplitude features. The raw amplitude data from multiple times are simultaneously input into the second encoder to encode second raw amplitude features. The multiple first and second raw amplitude features are input into the feature interaction module to interact and generate a fused amplitude feature. The fused amplitude feature is input into the decoder to decode a target amplitude feature. The target amplitude feature is input into the multi-task processing module to generate multiple amplitude task information. This embodiment performs feature interaction and fusion between the encoder and decoder, which can increase the information content of the features, thereby improving the robustness of the amplitude processing network and the accuracy of multi-task processing. Furthermore, the adaptive use of amplitude features to perform multi-task processing provides high flexibility, thus improving processing efficiency.

[0172] Example 3

[0173] See Figure 4 The diagram shows a schematic representation of the structure of a training device for an amplitude processing network according to Embodiment 3 of the present invention. Figure 4 As shown, the device includes:

[0174] The amplitude processing network loading module 401 is used to load the amplitude processing network when the original amplitude data of the pumped storage unit is collected at multiple times; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder and a multi-task processing module.

[0175] The first original amplitude feature encoding module 402 is used to input the original amplitude data at each time moment into the first encoder to encode the first original amplitude feature;

[0176] The second original amplitude feature encoding module 403 is used to simultaneously input the original amplitude data at multiple times into the second encoder to encode the second original amplitude feature;

[0177] The fusion amplitude feature interaction module 404 is used to input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and form fusion amplitude features;

[0178] The target amplitude feature decoding module 405 is used to input the fused amplitude feature into the decoder and decode it into the target amplitude feature;

[0179] Amplitude task information generation module 406 is used to input the target amplitude feature into the multi-task processing module to generate multiple amplitude task information;

[0180] The total loss value generation module 407 is used to generate a total loss value based on multiple amplitude task information.

[0181] The amplitude processing network update module 408 is used to update the amplitude processing network based on the total loss value.

[0182] In one embodiment of the present invention, the fused amplitude feature interaction module 404 includes:

[0183] The first candidate amplitude feature generation module is used to perform a normalization operation on each of the first original amplitude features to obtain the first candidate amplitude features;

[0184] The second candidate amplitude feature generation module is used to perform a normalization operation on the second original amplitude feature to obtain the second candidate amplitude feature;

[0185] The attention weight calculation module is used to calculate the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature;

[0186] The fusion amplitude feature alignment module is used to align the second candidate vibration feature into a fusion amplitude feature based on the attention weight.

[0187] In one embodiment of the present invention, the first candidate amplitude feature is represented as:

[0188]

[0189] Where i is time, C1 is the first original amplitude feature, U1 is the average value of the first original amplitude feature, σ1 is the variance of the first original amplitude feature, ε is the hyperparameter, and S1 is the first candidate amplitude feature.

[0190] The second candidate amplitude feature is represented as follows:

[0191]

[0192] Wherein, C2 is the second original amplitude feature, U2 is the average value of the second original amplitude feature, σ2 is the variance of the second original amplitude feature, ε is the hyperparameter, and S2 is the second candidate amplitude feature.

[0193] In one embodiment of the present invention, the attention weight calculation module includes:

[0194] The benchmark weight calculation module is used to calculate multiple benchmark weights in the attention mechanism, using the first candidate amplitude feature as the key value vector and the second candidate amplitude feature as the query vector.

[0195] The average value alignment module is used to add the product between the average value of the first original amplitude feature and the reference weight to obtain the aligned average value;

[0196] The average value weighting calculation module is used to linearly fuse the average value of the second original amplitude feature with the aligned average value to form the feature interaction average value;

[0197] The variance alignment module is used to add the product between the variance of the first original amplitude feature and the benchmark weight to obtain the aligned variance;

[0198] The variance weighting calculation module is used to linearly fuse the variance of the second original amplitude feature with the aligned variance into a feature interaction variance;

[0199] The fused amplitude characteristic is represented as follows:

[0200] C3 = S2 × Σ + U;

[0201] Wherein, C3 is the fused amplitude feature, S2 is the second candidate vibration feature, U is the feature interaction average, and Σ is the feature interaction variance.

[0202] In one embodiment of the present invention, the multi-task processing module includes a sequence prediction module, a working condition classification module and an anomaly discriminator, and the multiple amplitude task information includes target amplitude data, target working condition status and anomaly discrimination label;

[0203] The amplitude task information generation module 406 includes:

[0204] The target amplitude data generation module is used to input the target amplitude features into the sequence prediction module to generate the target amplitude data for the next time step;

[0205] The target operating condition classification module is used to input the target amplitude feature into the operating condition classification module to classify the pumped storage unit into a target operating condition.

[0206] An anomaly identification label generation module is used to input the target amplitude data into the anomaly detector to generate anomaly identification labels; the anomaly identification labels indicate whether the target amplitude data is abnormal.

[0207] In one embodiment of the present invention, the total loss value generation module 407 includes:

[0208] The first sub-loss value generation module is used to generate a first sub-loss value based on the difference between the target amplitude data and the original amplitude data at the next moment;

[0209] The true / false label generation module is used to input the target amplitude data into a preset true / false discriminator to generate true / false labels; the true / false labels indicate whether the target amplitude data is true or false.

[0210] The second sub-loss value generation module is used to generate a second sub-loss value based on the true / false discrimination label;

[0211] The third sub-loss value generation module is used to generate a third sub-loss value based on the difference between the target operating condition and the sample operating condition labeled for the pumped storage unit.

[0212] The fourth sub-loss value generation module is used to generate a fourth sub-loss value based on the anomaly discrimination label;

[0213] The total loss value fusion module is used to fuse the first sub-loss value, the second sub-loss value, the third sub-loss value and the fourth sub-loss value into a total loss value.

[0214] In one embodiment of the present invention, the first encoder is a backbone network with a self-attention mechanism, the backbone network with the self-attention mechanism includes a neural network Transformer, the second encoder is a long short-term memory network, and the decoder is a long short-term memory network.

[0215] The sequence prediction module consists of multiple fully connected layers, the working condition classification module consists of multiple fully connected layers, and the anomaly discriminator consists of multiple fully connected layers.

[0216] The training device for the amplitude processing network provided in the embodiments of the present invention can execute the training method for the amplitude processing network provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the training method for the amplitude processing network.

[0217] Example 4

[0218] See Figure 5 The diagram shows a structural schematic of an amplitude processing device provided in Embodiment 4 of the present invention. Figure 5 As shown, the device includes:

[0219] The amplitude processing network loading module 501 is used to load the amplitude processing network trained according to the method described in Embodiment 1 when the original amplitude data of the pumped storage unit is collected at multiple times; the amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder and a multi-task processing module.

[0220] The first original amplitude feature encoding module 502 is used to input the original amplitude data at each time moment into the first encoder to encode the first original amplitude feature;

[0221] The second original amplitude feature encoding module 503 is used to simultaneously input the original amplitude data at multiple times into the second encoder to encode the second original amplitude feature;

[0222] The fusion amplitude feature interaction module 504 is used to input multiple first original amplitude features and second original amplitude features into the feature interaction module to interact and form fusion amplitude features;

[0223] The target amplitude feature decoding module 505 is used to input the fused amplitude feature into the decoder and decode it into the target amplitude feature;

[0224] The amplitude task information generation module 506 is used to input the target amplitude feature into the multi-task processing module to generate multiple amplitude task information.

[0225] In one embodiment of the present invention, the fused amplitude feature interaction module 504 includes:

[0226] The first candidate amplitude feature generation module is used to perform a normalization operation on each of the first original amplitude features to obtain the first candidate amplitude features;

[0227] The second candidate amplitude feature generation module is used to perform a normalization operation on the second original amplitude feature to obtain the second candidate amplitude feature;

[0228] The attention weight calculation module is used to calculate the attention weight of the first candidate amplitude feature relative to the second candidate amplitude feature;

[0229] The fusion amplitude feature alignment module is used to align the second candidate vibration feature into a fusion amplitude feature based on the attention weight.

[0230] In one embodiment of the present invention, the first candidate amplitude feature is represented as:

[0231]

[0232] Where i is time, C1 is the first original amplitude feature, U1 is the average value of the first original amplitude feature, σ1 is the variance of the first original amplitude feature, ε is the hyperparameter, and S1 is the first candidate amplitude feature.

[0233] The second candidate amplitude feature is represented as follows:

[0234]

[0235] Wherein, C2 is the second original amplitude feature, U2 is the average value of the second original amplitude feature, σ2 is the variance of the second original amplitude feature, ε is the hyperparameter, and S2 is the second candidate amplitude feature.

[0236] In one embodiment of the present invention, the attention weight calculation module includes:

[0237] The benchmark weight calculation module is used to calculate multiple benchmark weights in the attention mechanism, using the first candidate amplitude feature as the key value vector and the second candidate amplitude feature as the query vector.

[0238] The average value alignment module is used to add the product between the average value of the first original amplitude feature and the reference weight to obtain the aligned average value;

[0239] The average value weighting calculation module is used to linearly fuse the average value of the second original amplitude feature with the aligned average value to form the feature interaction average value;

[0240] The variance alignment module is used to add the product between the variance of the first original amplitude feature and the benchmark weight to obtain the aligned variance;

[0241] The variance weighting calculation module is used to linearly fuse the variance of the second original amplitude feature with the aligned variance into a feature interaction variance;

[0242] The fused amplitude characteristic is represented as follows:

[0243] C3 = S2 × Σ + U;

[0244] Wherein, C3 is the fused amplitude feature, S2 is the second candidate vibration feature, U is the feature interaction average, and Σ is the feature interaction variance.

[0245] In one embodiment of the present invention, the multi-task processing module includes a sequence prediction module, a working condition classification module and an anomaly discriminator, and the multiple amplitude task information includes target amplitude data, target working condition status and anomaly discrimination label;

[0246] The amplitude task information generation module 506 includes:

[0247] The target amplitude data generation module is used to input the target amplitude features into the sequence prediction module to generate the target amplitude data for the next time step;

[0248] The target operating condition classification module is used to input the target amplitude feature into the operating condition classification module to classify the pumped storage unit into a target operating condition.

[0249] An anomaly identification label generation module is used to input the target amplitude data into the anomaly detector to generate anomaly identification labels; the anomaly identification labels indicate whether the target amplitude data is abnormal.

[0250] In one embodiment of the present invention, the first encoder is a backbone network with a self-attention mechanism, the backbone network with the self-attention mechanism includes a neural network Transformer, the second encoder is a long short-term memory network, and the decoder is a long short-term memory network.

[0251] The sequence prediction module consists of multiple fully connected layers, the working condition classification module consists of multiple fully connected layers, and the anomaly discriminator consists of multiple fully connected layers.

[0252] The amplitude processing device provided in the embodiments of the present invention can execute the amplitude processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the amplitude processing method.

[0253] Example 5

[0254] See Figure 6 This diagram illustrates a structural schematic of an electronic device according to an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0255] like Figure 6As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0256] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0257] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as training methods for amplitude processing networks or amplitude processing methods.

[0258] In some embodiments, the training method or amplitude processing method of the amplitude processing network may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the training method or amplitude processing method of the amplitude processing network described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the training method or amplitude processing method of the amplitude processing network by any other suitable means (e.g., by means of firmware).

[0259] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0260] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0261] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0262] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0263] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0264] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0265] Example 6

[0266] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements a training method or amplitude processing method for an amplitude processing network as provided in any embodiment of this invention.

[0267] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0268] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0269] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method of training an amplitude processing network, characterized by, The method comprises the following steps: loading an amplitude processing network when collecting original amplitude data of pumped storage units at multiple time points; the amplitude processing network comprises a first encoder, a second encoder, a feature interaction module, a decoder and a multi-task processing module; inputting the original amplitude data at each time point into the first encoder to encode first original amplitude features; inputting the original amplitude data at multiple time points into the second encoder to encode second original amplitude features; inputting the first original amplitude features and the second original amplitude features into the feature interaction module to interact into fused amplitude features; inputting the fused amplitude features into the decoder to decode into target amplitude features; inputting the target amplitude features into the multi-task processing module to generate multiple amplitude task information; generating a total loss value according to the multiple amplitude task information; updating the amplitude processing network according to the total loss value; wherein the step of inputting the first original amplitude features and the second original amplitude features into the feature interaction module to interact into fused amplitude features comprises: performing a normalization operation on each first original amplitude feature to obtain a first candidate amplitude feature; the first candidate amplitude feature is represented as: wherein i is a time point, C1 is the first original amplitude feature, U1 is the mean value of the first original amplitude feature, σ1 is the variance of the first original amplitude feature, ε is a hyperparameter, and S1 is the first candidate amplitude feature; performing a normalization operation on the second original amplitude feature to obtain a second candidate amplitude feature; the second candidate amplitude feature is represented as: wherein C2 is the second original amplitude feature, U2 is the mean value of the second original amplitude feature, σ2 is the variance of the second original amplitude feature, ε is a hyperparameter, and S2 is the second candidate amplitude feature; in the attention mechanism, taking the first candidate amplitude feature as a key vector and the second candidate amplitude feature as a query vector to calculate multiple reference weights; adding the product of the mean value of the first original amplitude feature and the reference weight to obtain an aligned mean value; linearly fusing the mean value of the second original amplitude feature and the aligned mean value into a feature interaction mean value; adding the product of the variance of the first original amplitude feature and the reference weight to obtain an aligned variance; linearly fusing the variance of the second original amplitude feature and the aligned variance into a feature interaction variance; aligning the second candidate amplitude feature into a fused amplitude feature according to the attention weight; the fused amplitude feature is represented as: C3=S2×Σ+U; wherein C3 is the fused amplitude feature, S2 is the second candidate amplitude feature, U is the feature interaction mean value, and Σ is the feature interaction variance.

2. The method of claim 1, wherein, The multi-task processing module comprises a sequence prediction module, a working condition classification module and an anomaly discriminator, and the multiple amplitude task information comprises target amplitude data, a target working condition state and an anomaly discrimination label; the step of inputting the target amplitude features into the multi-task processing module to generate multiple amplitude task information comprises: inputting the target amplitude feature into the sequence prediction module to generate target amplitude data of a next time point; inputting the target amplitude feature into the working condition classification module to classify the pumped storage unit into a target working condition state; inputting the target amplitude data into the anomaly discriminator to generate an anomaly discrimination label; the anomaly discrimination label indicates whether the target amplitude data is abnormal.

3. The method of claim 2, wherein, The total loss value is generated according to the plurality of amplitude task information, including: a first sub-loss value is generated according to the difference between the target amplitude data and the original amplitude data of the next time point; a true-false discrimination label is generated by inputting the target amplitude data into a preset true-false discriminator; the true-false discrimination label indicates whether the target amplitude data is true or false; a second sub-loss value is generated according to the true-false discrimination label; a third sub-loss value is generated according to the difference between the target working condition state and a sample working condition state labeled for the pumped storage unit; a fourth sub-loss value is generated according to the anomaly discrimination label; The first sub-loss value, the second sub-loss value, the third sub-loss value and the fourth sub-loss value are fused into a total loss value.

4. The method of claim 2, wherein, The first encoder is a backbone network applying a self-attention mechanism, the backbone network applying the self-attention mechanism includes a neural network Transformer, the second encoder is a long short-term memory network, and the decoder is a long short-term memory network; The sequence prediction module is a plurality of fully connected layers, the working condition classification module is a plurality of fully connected layers, and the anomaly discriminator is a plurality of fully connected layers.

5. An amplitude processing method, characterized by, including: When original amplitude data of a pumped storage unit is collected at a plurality of time points, a vibration amplitude processing network trained by the method according to any one of claims 1-4 is loaded; The vibration amplitude processing network includes a first encoder, a second encoder, a feature interaction module, a decoder and a multi-task processing module; The original amplitude data of each time point is respectively input into the first encoder to encode a first original amplitude feature; The original amplitude data of a plurality of time points is simultaneously input into the second encoder to encode a second original amplitude feature; The first original amplitude features and the second original amplitude feature are input into the feature interaction module to interact into a fusion amplitude feature; The fusion amplitude feature is input into the decoder to decode into a target amplitude feature; The target amplitude feature is input into the multi-task processing module to generate a plurality of amplitude task information.

6. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected in communication with the at least one processor; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the training method of the vibration amplitude processing network according to any one of claims 1-4 or the vibration amplitude processing method according to claim 5.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the training method of the vibration amplitude processing network according to any one of claims 1-4 or the vibration amplitude processing method according to claim 5.

Images