Intelligent control method and system for switching power supply

Abstract

The invention discloses a switching power supply intelligent control method and system, and relates to the technical field of switching power supply intelligent control, and the method comprises the steps: obtaining the operation data of equipment, and determining the average operation state voiceprint of the equipment through the operation data of the equipment; based on the average operation state voiceprint of the plurality of equipment samples and the fault prediction result of the equipment, training a neural network to obtain an equipment fault prediction model; and inputting the average operation state voiceprint of the equipment to the equipment fault prediction model in real time to obtain an equipment fault prediction result output by the equipment fault prediction model. According to the method, the equipment operation state is better represented through the first average operation state voiceprint, the second average operation state voiceprint and the third average operation state voiceprint, and the current parameter, the voltage parameter, the temperature parameter, the environment parameter and the management measure parameter are used as input indexes; by means of the long-short-term memory network model, the prediction process of the equipment fault result is simplified, and meanwhile the prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of switching power supplies, and more specifically, to an intelligent control method and system for switching power supplies. Background Art

[0002] Early equipment fault prediction models were mainly data-driven. Some machine learning models in the context of data-driven, such as distributed gradient boosting models, random forest models, support vector machine models, and artificial neural networks, have been widely applied to equipment fault prediction applications.

[0003] However, the accuracy of empirical statistical models for equipment fault prediction depends on the data quality of a large number of training samples and the representativeness of the models. The accuracy of equipment fault prediction is always limited and the estimation effect is not good. In the actual popularization and application of equipment fault prediction, the method relying on a large number of observation samples consumes a large amount of manpower and material resources and is difficult to be applied on a large scale. Further, with the continuous in-depth understanding of the operating state of the equipment, analyzing the operating state is more conducive to equipment fault prediction, and the prediction accuracy is relatively better than that of empirical statistical models.

[0004] Therefore, how to propose an intelligent control method and system for switching power supplies to solve the problem of poor equipment fault prediction effect in the prior art and improve the intelligent control effect of switching power supplies is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent control method and system for switching power supplies to solve the problem of poor equipment fault prediction effect in the prior art and improve the intelligent control effect of switching power supplies. To achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent control method for a switching power supply, comprising:

[0007] Obtaining the operating data of the equipment, and determining the first average operating state soundprint, the second average operating state soundprint, and the third average operating state soundprint of the equipment through the operating data of the equipment;

[0008] Training a neural network based on the first average operating state soundprint, the second average operating state soundprint, the third average operating state soundprint of multiple equipment samples, and the equipment fault prediction results to obtain an equipment fault prediction model;

[0009] Real-time inputting the first average operating state soundprint, the second average operating state soundprint, and the third average operating state soundprint of the equipment into the equipment fault prediction model to obtain the equipment fault prediction results output by the equipment fault prediction model.

[0010] Optionally, the first average operating state voiceprint is the average operating state voiceprint from the startup to the stable operation stage, the second average operating state voiceprint is the average operating state voiceprint of a time series segment intercepted from the stable operation stage, and the third average operating state voiceprint is the average operating state voiceprint from the stable operation stage to shutdown.

[0011] Optionally, the first average operating state voiceprint, the second average operating state voiceprint, the third average operating state voiceprint of the multiple device samples, and the device fault prediction result are obtained by simulating multiple different scenarios through a device fault prediction model. Each scenario corresponds to a set of scenario parameter combinations, and the scenario parameter combinations include combinations of current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters.

[0012] Optionally, the input of the device fault prediction model is current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters, and the output is the fault prediction result.

[0013] Optionally, obtaining the operation data of the device includes:

[0014] Determining the startup type of the device according to the startup method of the device;

[0015] According to the startup type of the device, determining the first time range from startup to the stable operation stage, the second time range of the stable operation stage, and the third time range from the stable operation stage to shutdown of the device; obtaining the corresponding operation data based on the first time range, the second time range, and the third time range.

[0016] Optionally, the neural network is a long short-term memory network.

[0017] Optionally, the method for determining the average operating state voiceprint includes:

[0018] Using the average operating state signal formed by current parameters, voltage parameters, temperature parameters, and environmental parameters to indicate the device state in a certain time period, collecting the intensity magnitudes of the operating state signals at different time points within the determined time range, and representing the intensity of the operating state signal at the measured time point measured by the parameter type as: {φ ij (t), t = 1, …, k (k ≥ 1)}; where φ ij (t) represents the intensity of the operating state signal measured at the t-th time, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th parameter type, and the intensity of the operating state signal at the i-th measured time point is obtained through j parameter types to construct the average operating state voiceprint library ψ.

[0019] Optionally, the device fault prediction model includes:

[0020] The forget gate reads the output information h at the previous moment t-1 and the input information x at the current moment t and outputs a value f between 0 and 1 t For each number in the cell state, 1 means completely retain, 0 means completely discard;

[0021] The sigmoid layer of the input gate determines the value i to be updated t and creates a new candidate cell state vector through the tanh layer Multiply the old cell state C t-1 by the f obtained from the forget gate t to discard the information that needs to be discarded, and then add the updated value i obtained from the input gate t multiplied by the new cell candidate value vector to obtain the new cell state C t ;

[0022] The input information x at the current moment t will pass through a sigmoid layer to determine the result o to be output t Then, multiply the result o to be output t by the cell state C processed by the tanh layer t to obtain the determined output h at the current moment t .

[0023] Optionally, it further includes: verifying the estimation accuracy of the equipment fault prediction model, including verifying with measured sample points. The data of the measured sample points verification comes from the sample data of different factory areas. The accuracy verification indicators include the coefficient of determination R 2 , root mean square error RMSE and mean relative error MRE. The higher the R 2 , the smaller the RMSE and MRE, indicating the better performance of the equipment fault prediction model.

[0024] Optionally, a smart control system for a switching power supply includes:

[0025] Acquisition module: used to obtain the operation data of the equipment;

[0026] Preprocessing module: used to determine the first average operation state soundprint, the second average operation state soundprint and the third average operation state soundprint of the equipment through the operation data of the equipment;

[0027] Model construction module: used to train a recurrent neural network based on the first average operation state soundprint, the second average operation state soundprint, the third average operation state soundprint of multiple equipment samples and the equipment fault prediction results to obtain an equipment fault prediction model;

[0028] Prediction module: It is used to input the first average operating state voiceprint, the second average operating state voiceprint, and the third average operating state voiceprint of the device in real time into the device fault prediction model, and obtain the fault prediction result of the device output by the device fault prediction model.

[0029] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method and system for intelligent management and control of a switching power supply, which has the following beneficial effects:

[0030] The present invention proposes a method for intelligent management and control of a switching power supply, including: obtaining the operating data of the device, and determining the first average operating state voiceprint, the second average operating state voiceprint, and the third average operating state voiceprint of the device through the operating data of the device; training a neural network based on the first average operating state voiceprint, the second average operating state voiceprint, the third average operating state voiceprint of multiple device samples, and the fault prediction result of the device to obtain a device fault prediction model; inputting the first average operating state voiceprint, the second average operating state voiceprint, and the third average operating state voiceprint of the device into the device fault prediction model in real time, and obtaining the fault prediction result of the device output by the device fault prediction model. The present invention solves the problem of poor prediction effect of device fault results in the prior art, better characterizes the device operating state through the first average operating state voiceprint, the second average operating state voiceprint, and the third average operating state voiceprint, simplifies the prediction process of device fault results by using current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters as input indicators, and improves the prediction accuracy at the same time. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0032] Figure 1 It is a schematic flowchart of a method for intelligent management and control of a switching power supply provided by the present invention.

[0033] Figure 2 It is a structural framework diagram of a system for intelligent management and control of a switching power supply provided by the present invention. Detailed Embodiments

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] An embodiment of the present invention discloses an intelligent control method for a switching power supply, as Figure 1 shown, including:

[0036] Obtain the operation data of the device, and determine the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device through the operation data of the device;

[0037] Based on the first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint of multiple device samples, and the fault prediction results of the device, train a neural network to obtain a device fault prediction model;

[0038] Input the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device into the device fault prediction model in real time, and obtain the fault prediction results of the device output by the device fault prediction model.

[0039] Further, the first average operation state voiceprint is the average operation state voiceprint from the start to the stable operation stage, the second average operation state voiceprint is the average operation state voiceprint of the time series segment intercepted from the stable operation stage, and the third average operation state voiceprint is the average operation state voiceprint from the stable operation stage to shutdown.

[0040] Further, the first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint of the multiple device samples, and the fault prediction results of the device are obtained by simulating a variety of different scenarios through the device fault prediction model. Each scenario corresponds to a set of scenario parameter combinations, and the scenario parameter combinations include combinations of current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters.

[0041] Further, the inputs of the device fault prediction model are current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters, and the output is the fault prediction result. Among them, the management measure parameters include the detection date, etc.

[0042] Further, the obtaining of the operation data of the device includes:

[0043] Determine the start type of the device according to the start mode of the device;

[0044] Determine a first time range from when the device is turned on to the stable operation stage, a second time range of the stable operation stage, and a third time range from the stable operation stage to shutdown according to the startup type of the device; obtain corresponding operation data based on the first time range, the second time range, and the third time range.

[0045] Further, determining the startup type of the device according to the startup mode of the device includes: determining a classification set of different devices according to the startup mode of the device (such as star-delta startup, autotransformer step-down startup, or soft startup) and corresponding load parameters. Determine the first time range from when the device is turned on to the stable operation stage, the second time range of the stable operation stage, and the third time range from the stable operation stage to shutdown according to the classification set of different devices.

[0046] Specifically, the second average operation state voiceprint is the average operation state voiceprint of the time series segment intercepted from the stable operation stage, including intercepting a time segment with a length of (the first time range + the third time range) / 2 from the stable operation stage as the second time range, and determining the second average operation state voiceprint through the second time range.

[0047] Further, the neural network is a long short-term memory network.

[0048] Further, the method for determining the average operation state voiceprint includes: using the average operation state signal formed by current parameters, voltage parameters, temperature parameters, and environmental parameters to indicate the device state in a certain time period, collecting the intensity magnitudes of the operation state signals at different time points within the determined time range, and representing the intensity of the operation state signal at the measured time point measured by the parameter type as: {φ ij (t), t = 1, …, k (k ≥ 1)}; where φ ij (t) represents the intensity of the operation state signal measured at the t-th measurement, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th parameter type, and the intensity of the operation state signal at the i-th measured time point is obtained through j parameter types to construct an average operation state voiceprint library ψ.

[0049] In the specific implementation manner, the method for determining the average operation state voiceprint specifically includes:

[0050] Using the average operation state signal formed by current parameters, voltage parameters, temperature parameters, and environmental parameters to indicate the device state in a certain time period, collecting the intensity magnitudes of the operation state signals at different time points within the determined time range, and representing the intensity of the operation state signal at the measured time point measured by the parameter type as:

[0051] {φ ij(t), where t = 1, …, k (k ≥ 1)};

[0052] Among them, φ ij (t) represents the running state signal strength of the t-th measurement, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th parameter type, and the running state signal strength of the i-th measured time point is obtained through j parameter types to construct the average running state fingerprint library ψ.

[0053]

[0054] Among them, m is the number of parameter types, n is the number of groups of time points to be measured, and the i-th row vector ψ i = [φ i1 , φ i2 ,..., φ im is the fingerprint feature value of the measured time point measured by m parameter types. Different types of parameters are used to measure different feature quantities to form a multi-dimensional feature quantity dimension. When a device failure occurs at the measured time point, the information measured by each parameter type is the running state signal strength.

[0055] In the specific implementation manner, the above method for determining the average running state voiceprint is adopted. According to the startup type of the device, the first time range from the start to the stable operation stage, the second time range of the stable operation stage, and the third time range from the stable operation stage to shutdown of the device are determined; based on the first time range, the second time range, and the third time range, the corresponding running data is obtained, and the first average running state voiceprint, the second average running state voiceprint, and the third average running state voiceprint of the device are determined according to the running data of the above device.

[0056] Further, the device failure prediction model includes:

[0057] The forget gate reads the output information h t-1 of the previous moment and the input information x t of the current moment, and outputs a value f t between 0 and 1 to each number in the cell state. 1 means completely retain, and 0 means completely discard;

[0058] The sigmoid layer of the input gate determines the value i t to be updated, and a new cell state candidate value vector is created through the tanh layer Multiply the old cell state C t-1 by the f t obtained by the forget gate, discard the information that needs to be discarded, and then add the updated value i t obtained by the input gate and the new cell candidate value vector The product is taken to obtain a new cell state C t ;

[0059] The input information x at the current moment t will pass through a sigmoid layer to determine the result o to be output t , then, the result o to be output t is multiplied by the cell state processed by the tanh layer C t to obtain the output h determined at the current moment t .

[0060] In a specific embodiment, the long short-term memory network (LSTM) is a machine learning algorithm widely used in time series signal prediction and classification. LSTM is a special type of RNN that can learn long-term dependence information. The basic structure is the same as that of RNN, which is a chain form of repeating neural network modules. LSTM has four special layers for interaction between input and output.

[0061] The LSTM network is like a continuously updated cell, and its key concept is the cell state. LSTM has the ability to remove or add information to the cell state through a carefully designed structure called a "gate". The gate allows information to pass selectively. 0 represents not allowing any amount to pass, and 1 represents allowing any amount to pass. LSTM has three gates to update the cell state, namely the forget gate, the input gate, and the output gate. Among them, the forget gate and the input gate are the main parts that can keep the network active in the long short-term memory network. As the network is continuously iteratively updated, they determine the information to be discarded and retained in the network. Specifically:

[0062] First step, determine the information to be discarded from the cell state. This is completed through the sigmoid layer of the "forget gate". The forget gate reads the output information h at the previous moment t-1 and the input information x at the current moment t , and outputs a value f between 0 and 1 t for each number in the cell state. 1 means completely retain, and 0 means completely discard.

[0063] f t =σ(W f ·[h t-1 , x t +b f );

[0064] Second step, determine the new information to be stored in the cell state. First, the sigmoid layer of the "input gate" determines the value i to be updated t. Second, create a new candidate value vector for the cell state through the tanh layer

[0065] i t = σ(W f · [h t-1 , x t + b i );

[0066]

[0067] Multiply the old cell state C t-1 by the f t obtained from the "forget gate" to discard the information that needs to be discarded, and then add the product of the update value i t obtained from the "input gate" and the new cell candidate value vector to get the new cell state C t ,

[0068]

[0069] Thirdly, it is necessary to determine the output result (prediction result). The output result will be based on the current cell state C t . The input information x t at the current moment will pass through a sigmoid layer to determine the result o t to be output. Then, multiply the result o t to be output by the cell state processed by the tanh layer C t to get the output h t determined at the current moment.

[0070] o t = σ(W o · [h t-1 , x t + b o );

[0071] h t = o t * tanh(C t );

[0072] The input of the LSTM network is current parameter, voltage parameter, temperature parameter and environmental parameter data, defined as X t = [x t RH , x t RK , x t RA , x t LH , and the output is the prediction result Y t+1 = [yt+1 RH , y t+1 RK , y t+1 RA , y t+1 LH ]。

[0073] To evaluate the performance of the LSTM network for equipment fault prediction, the root mean square error RMSE (E RMS ) is used to quantify the difference between the predicted value and the actual observed value.

[0074]

[0075] In the specific implementation, the LSTM model is built in the MATLAB environment. The LSTM network architecture is defined as follows: the number of input features and the number of output responses are both 4, and the number of hidden units in the network is set to 200. The solver for training the model is set to 'Adam' and trained for 250 rounds. To prevent gradient explosion, the gradient threshold is set to 1, the initial learning rate is 0.005, and the learning rate is reduced by multiplying by the reduction factor 0.5 after every 50 rounds of training. The dataset is from equipment operation data, including current parameter (RH), voltage parameter (RK), temperature parameter (RA), and environmental parameter data (LH). To prevent training divergence, the training data needs to be standardized to a dataset with zero mean and unit variance first. The dataset is divided into a training dataset and a test dataset in a ratio of 6:4. The goal is to use past historical equipment data to predict the equipment fault result for the next time step, and the prediction time step is defined as 1 sampling point. The historical equipment data vector is used as the input vector of the LSTM network, and the predicted equipment fault result is the output vector of the LSTM.

[0076] In the specific implementation, during the process of training the model, the model optimizer is Adam, the loss function is mean squared error, and the network is trained multiple times to find the optimal values of parameters such as the number of model layers, the number of neurons units, epochs, batch size, dropout, etc.

[0077] Further, it also includes: verifying the estimation accuracy of the equipment fault prediction model, including verifying with measured sample points. The data for verifying with measured sample points comes from sample data of different factory areas. The accuracy verification indicators include the coefficient of determination R 2 , the root mean square error RMSE, and the mean relative error MRE. The higher the R 2 , and the smaller the RMSE and MRE, the better the performance of the equipment fault prediction model.

[0078]

[0079] where x i and y i represent the actual prediction value and the estimated prediction value respectively, is the mean value of the actual prediction values.

[0080] In a specific embodiment, an intelligent control system for a switching power supply, as Figure 2 shown, includes:

[0081] Acquisition module: used to obtain the operation data of the device;

[0082] Preprocessing module: used to determine the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device through the operation data of the device;

[0083] Model construction module: a device fault prediction model obtained by training a recurrent neural network based on the first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint of multiple device samples, and the device fault prediction results;

[0084] Prediction module: used to input the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device into the device fault prediction model in real time, and obtain the device fault prediction results output by the device fault prediction model.

[0085] In a specific embodiment, there is also provided an electronic device, which includes: a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus. The processor can call the logical instructions in the memory to execute an intelligent control method for a switching power supply. In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0086] In a specific embodiment, it further includes a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements a method for intelligent control of a switching power supply provided by the above-mentioned various methods.

[0087] In a specific embodiment, it further includes a computer program product. The computer program product includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a method for intelligent control of a switching power supply provided by the above-mentioned various methods.

[0088] In the present specification, the various embodiments are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the description in the method part.

[0089] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent control method for a switching power supply, characterized in that, Including: Obtain the operation data of the device, and determine the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device based on the operation data of the device; Train a neural network based on the first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint of multiple device samples, and the device fault prediction result to obtain a device fault prediction model; Input the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device into the device fault prediction model in real time to obtain the device fault prediction result output by the device fault prediction model.

2. The intelligent control method for a switching power supply according to claim 1, wherein The first average operation state voiceprint is the average operation state voiceprint from startup to stable operation stage, the second average operation state voiceprint is the average operation state voiceprint of the time series segment intercepted from the stable operation stage, and the third average operation state voiceprint is the average operation state voiceprint from the stable operation stage to shutdown.

3. The intelligent control method of a switching power supply according to claim 1, wherein The first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint, and the device fault prediction result of the multiple device samples are obtained by simulating various different scenarios through a device fault prediction model. Each scenario corresponds to a set of scenario parameter combinations, and the scenario parameter combinations include combinations of current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters.

4. The intelligent control method for a switching power supply according to claim 1, wherein The input of the device fault prediction model is current parameters, voltage parameters, temperature parameters, environmental parameters, and management measure parameters, and the output is the fault prediction result.

5. The intelligent control method of a switching power supply according to claim 1, wherein The obtaining of the operation data of the device includes: Determine the startup type of the device according to the startup method of the device; According to the startup type of the device, determine the first time range from startup to stable operation stage, the second time range of the stable operation stage, and the third time range from the stable operation stage to shutdown of the device; obtain the corresponding operation data based on the first time range, the second time range, and the third time range.

6. The intelligent control method for a switching power supply according to claim 1, wherein The neural network is a long short-term memory network.

7. The intelligent control method of a switching power supply according to claim 2, characterized in that, The method for determining the average operation state voiceprint includes: The average operating state signal formed by current parameters, voltage parameters, temperature parameters, and environmental parameters indicates the equipment state in a certain time period. The intensity of the operating state signals at different time points within a determined time range is collected. The intensity of the operating state signal at the measured time point measured by the parameter type is expressed as: {φ ij (t), t = 1, …, k (k ≥ 1)}; where φ ij (t) represents the intensity of the operating state signal at the t-th measurement, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th parameter type, and the intensity of the operating state signal at the i-th measured time point is obtained through j parameter types to construct the average operating state voiceprint library ψ.

8. The intelligent control method of a switching power supply according to claim 1, characterized in that The device fault prediction model includes: The forget gate reads the output information h at the previous moment t-1 and the input information x at the current moment t and outputs a value f between 0 and 1 t For each number in the cell state, 1 means complete retention and 0 means complete discard; The sigmoid layer of the input gate determines the value i to be updated t , and a new candidate value vector for the cell state is created through the tanh layer Multiply the old cell state C t-1 by f obtained from the forget gate t to discard the information that needs to be discarded, and then add the updated value i obtained from the input gate t multiplied by the new candidate value vector of the cell to obtain the new cell state C t ; The input information x at the current moment t will pass through a sigmoid layer to determine the result o to be output t , then, the result o to be output t is multiplied by the cell state C processed by the tanh layer t to obtain the output h determined at the current moment t .

9. The intelligent control method of a switching power supply according to claim 1, wherein Also including: Verify the estimation accuracy of the equipment failure prediction model, including verification with measured sample points. The verification data of measured sample points comes from sample data of different factory areas. The accuracy verification indicators include the coefficient of determination R 2 , root mean square error RMSE, and mean relative error MRE. The higher the R 2 , the smaller the RMSE and MRE, indicating that the performance of the equipment failure prediction model is better.

10. An intelligent control system for a switching power supply, characterized in that, Including: Acquisition module: used to obtain the operation data of the device; Preprocessing module: used to determine the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device based on the operation data of the device; Model construction module: used to train a recurrent neural network based on the first average operation state voiceprint, the second average operation state voiceprint, the third average operation state voiceprint of multiple device samples, and the device fault prediction result to obtain a device fault prediction model; Prediction module: used to input the first average operation state voiceprint, the second average operation state voiceprint, and the third average operation state voiceprint of the device into the device fault prediction model in real time to obtain the device fault prediction result output by the device fault prediction model.

Images