Training data generation method, model training method, electronic equipment and medium

Abstract

The embodiment of the invention provides a training data generation method, and the method comprises the steps: obtaining the rotation angular displacement data of a main shaft of a circuit breaker according to a time sequence in each switching-off / switching-on process of multiple switching-on and switching-off of a contact of the circuit breaker; converting the rotation angular displacement data of the main shaft according to the time sequence into linear displacement data of a contact according to the time sequence to obtain first time sequence data; acquiring a control signal in the circuit breaker; converting the first time sequence data from analog quantity to digital quantity to obtain second time sequence data; performing filtering fusion processing on the second time sequence data by adopting a preset filtering fusion algorithm to obtain third time sequence data; calibrating the third time sequence data according to a control signal in the circuit breaker to obtain fourth time sequence data; and according to a preset classification rule, marking corresponding classification labels on the multiple groups of fourth time sequence data.

Description

Training data generation method, model training method, electronic device and medium Technical Field

[0001] The embodiments of the present application mainly relate to the field of artificial intelligence, and in particular to a training data generation method, a model training method, an electronic device, and a medium. Background Art

[0002] Switchgear is a type of electrical equipment whose primary function is to open and close, control, and protect electrical equipment during the power generation, transmission, distribution, and conversion processes of power systems. Medium-voltage switchgear is a key component in the medium-voltage power sector, and its condition is crucial to the safety of the entire power station. To monitor the status of medium-voltage switchgear, the travel trajectory of the circuit breaker contacts is typically measured online, compared to standard discrete characteristic values, and the test results are manually determined. However, this method not only makes it difficult to accurately measure the contact travel trajectory, but also results in highly subjective tests.

[0003] Summary of the Invention

[0004] The embodiments of the present application provide a training data generation method, a model training method, an electronic device, and a medium, which generate high-quality status data of circuit breakers in a medium-voltage switchgear to train a neural network model. The trained neural network model can quickly and accurately predict the status of the circuit breaker.

[0005] In a first aspect, a training data generation method is provided, comprising: obtaining rotational angular displacement data of a main shaft of a circuit breaker in a time series during each opening / closing process of multiple opening / closing of contacts of the circuit breaker; converting the rotational angular displacement data of the main shaft in a time series into linear displacement data of contacts in a time series to obtain first time series data; obtaining a control signal inside the circuit breaker; converting the first time series data from an analog quantity to a digital quantity to obtain second time series data; performing filtering and fusion processing on the second time series data using a preset filtering and fusion algorithm to obtain third time series data; calibrating the third time series data according to the control signal inside the circuit breaker to obtain fourth time series data; and assigning corresponding classification labels to multiple groups of fourth time series data according to preset classification rules.

[0006] In a second aspect, a neural network model training method is provided, comprising: executing the training data generation method provided in the first aspect to obtain a training data set containing a preset amount of training data; using the training data set to train the neural network model to obtain an optimized neural network model; wherein the neural network model comprises: a CNN model or an LSTM model.

[0007] In a third aspect, an electronic device is provided, comprising: at least one memory configured to store computer-readable code; and at least one processor configured to call the computer-readable code and execute each step of the method provided in the first aspect.

[0008] In a fourth aspect, a computer-readable medium is provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the processor executes each step in the method provided in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The following figures are intended only to illustrate and explain the embodiments of the present application and are not intended to limit the scope of the embodiments of the present application.

[0010] FIG1 is a flowchart of a method for generating training data according to an embodiment of the present application;

[0011] FIG2 is a schematic diagram of the internal structure of a circuit breaker according to an embodiment of the present application;

[0012] FIG3 is a flow chart of a neural network model training method according to an embodiment of the present application;

[0013] FIG4 is a schematic diagram of an electronic device according to an embodiment of the present application.

[0014] Description of Reference Numerals

[0015] 100: Training data generation method 101-107: Method steps

[0016] 20: Circuit breaker 21: Static contact 22: Moving contact

[0017] 23: Main Axis 300: Neural Network Model Training Method 301-302: Method Steps

[0018] 400: Electronic device 401: Processor 402: Memory DETAILED DESCRIPTION

[0019] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that discussing these embodiments is merely to enable those skilled in the art to better understand and implement the subject matter described herein, and is not intended to limit the scope of protection, applicability, or examples set forth in the claims. The functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the embodiments of the present application. Various examples may omit, replace, or add various processes or components as needed. For example, the described method may be performed in an order different from the described order, and various steps may be added, omitted, or combined. In addition, the features described relative to some examples may also be combined in other examples.

[0020] As used herein, the term "including" and its variations are open terms meaning "including but not limited to". The term "based on" means "based at least in part on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other definitions may be included below, whether explicit or implicit. Unless the context clearly indicates otherwise, the definition of a term is consistent throughout the specification.

[0021] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0022] FIG1 is a flow chart of a method for generating training data according to an embodiment of the present application. As shown in FIG1 , the method 100 for generating training data includes:

[0023] Step 101 : During each opening / closing process of multiple opening and closing operations of contacts of the circuit breaker, obtain rotational angular displacement data of a main shaft of the circuit breaker in a time series.

[0024] Alternatively, the rotational angular displacement data of the main shaft 23 in the four-bar linkage composed of L1, L2, L3 and L4 as shown in FIG2 may be acquired in a time series by a rotary encoder.

[0025] Step 102 : Convert the time-series rotational angular displacement data of the main shaft into time-series linear displacement data of the contactor to obtain first time-series data.

[0026] By calculating the four-bar linkage model in the circuit breaker 20, the model proportional coefficient k is obtained, and the rotation angle displacement data of the main shaft 23 in time series is converted to Converted into linear displacement data f of the contact in time series, The contact is composed of a static contact 21 and a moving contact 22 in FIG2 , wherein the static contact refers to the contact that does not move with the actuator, while the moving contact is the other way around.

[0027] Optionally, the first time series data may be stored in an array format or in a curve format. The storage format determines the data format that can be predicted by the trained neural network model.

[0028] Step 103: Acquire the control signal inside the circuit breaker.

[0029] The control signals within a circuit breaker include the current in the opening and closing coils and the signals that control the opening and closing of the contacts. The opening and closing coils are control components within the circuit control loop, and the opening and closing coil currents are specifically the coil currents in the electromagnetic mechanism that drives the circuit breaker's opening and closing operations.

[0030] Step 104 : Convert the first time series data from analog quantity to digital quantity to obtain second time series data.

[0031] Step 105 : Perform filtering and fusion processing on the second time series data using a preset filtering and fusion algorithm to obtain third time series data.

[0032] The preset filter fusion algorithm can be: Kalman filter algorithm or particle filter algorithm. Using the correlation filter fusion algorithm can reduce model estimation errors, such as errors in the f function caused by errors in related parameters, and related measurement errors.

[0033] Step 106: Calibrate the third time series data according to the control signal inside the circuit breaker to obtain fourth time series data.

[0034] The third time series data can be calibrated using the signal controlling the opening and closing of the contacts. Specifically, when the signal controlling the opening and closing of the contacts is 0, the corresponding third time series data is cleared. The third time series data can also be calibrated using the current signal of the opening and closing coils, for example, by deleting related outliers and aligning related time series data.

[0035] Step 107 : According to a preset classification rule, corresponding classification labels are respectively added to the plurality of groups of fourth time series data.

[0036] According to a preset classification rule, multiple sets of fourth time series data are divided into at least two categories. The fourth time series data in the first category is labeled as a first sample; the fourth time series data in the second category is labeled as a second sample. In one embodiment, positive and negative samples can be classified based on a preset number of segmentations. For example, data from the first 1000 segmentations can be classified as positive samples and labeled "normal," while data from the last 1000 segmentations can be classified as negative samples and labeled "abnormal." In one embodiment, samples whose average moving speed, segmentation time, or bounce interval exceeds a preset range can be classified as negative samples, while samples whose average moving speed, segmentation time, or bounce interval exceeds a preset range can be classified as positive samples. In one embodiment, multiple sets of fourth time series data can be classified into multiple categories based on their degree of aging. For example, data from the first 500 segmentations can be labeled as 50% aged, data from 500 to 1000 segmentations can be labeled as 70% aged, and data from 1000 to 1500 segmentations can be labeled as 90% aged.

[0037] Through the embodiment of the present application, the rotational angular displacement data of the main shaft in a time series is converted into the linear displacement data of the contact in a time series, and then through the means of relevant filtering fusion and relevant calibration, the relevant data are marked with corresponding classification labels, so that high-quality and closer to the real value state data for training the neural network model can be generated.

[0038] FIG3 is a flow chart of a neural network model training method according to an embodiment of the present application. As shown in FIG3 , the neural network model training method 300 includes:

[0039] Step 301 : executing the training data generating method 100 to obtain a training data set including a preset amount of training data.

[0040] Step 302: Use the training data set to train the neural network model to obtain an optimized neural network model. The neural network model includes: a CNN model or an LSTM model.

[0041] In one embodiment, the contact time series data of each opening / closing process of the circuit breaker in multiple online opening and closing operations is input into an optimized neural network model, and the optimized neural network model can be used to predict the corresponding label, for example, normal or abnormal.

[0042] The neural network model trained under the embodiment of the present application can quickly and accurately predict the status of the circuit breaker in the medium-voltage switchgear, thereby realizing intelligent monitoring and maintenance of the medium-voltage switchgear.

[0043] The present application also provides an electronic device 400. FIG4 is a schematic diagram of a device control platform 400 according to an embodiment of the present application. As shown in FIG4 , the device control platform 400 includes a processor 401 and a memory 402. The memory 402 stores instructions, wherein the instructions, when executed by the processor 401, implement the method 100 described above.

[0044] Among them, at least one processor 401 may include a microprocessor, an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a central processing unit (CPU), a graphics processing unit (GPU), a state machine, etc. Embodiments of computer-readable media include, but are not limited to, floppy disks, CD-ROMs, magnetic disks, memory chips, ROMs, RAMs, ASICs, configured processors, all-optical media, all tapes or other magnetic media, or any other medium from which a computer processor can read instructions. In addition, various other forms of computer-readable media can send or carry instructions to a computer, including routers, private or public networks, or other wired and wireless transmission devices or channels. Instructions may include code in any computer programming language, including C, C++, C language, Visual Basic, Java, and JavaScript.

[0045] In addition, the embodiments of the present application further provide a computer-readable medium having computer-readable instructions stored thereon, and the computer-readable instructions, when executed by the processor, cause the processor to execute the aforementioned training data generation method. Embodiments of computer-readable media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Optionally, computer-readable instructions can be downloaded from a server computer or a cloud via a communication network.

[0046] It should be noted that not all steps and modules in the above processes and system structure diagrams are required, and certain steps or modules can be omitted according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or may be implemented by certain components in multiple independent devices.

Claims

1. A training data generation method, characterized in that: include: - obtaining (101) the rotational angular displacement data of the main shaft of the circuit breaker in a time series during each opening / closing process of the contacts of the circuit breaker; - converting (102) the time-series rotational angular displacement data of the main shaft into time-series linear displacement data of the contact, thereby obtaining first time-series data; - obtaining (103) a control signal inside the circuit breaker; -Converting (104) the first time series data from analog quantity to digital quantity to obtain second time series data; - Using a preset filtering and fusion algorithm to perform filtering and fusion processing on the second time series data (105) to obtain third time series data; - performing calibration (106) on the third time series data according to the control signal inside the circuit breaker to obtain fourth time series data; - According to the preset classification rules, the multiple groups of fourth time series data are respectively marked with corresponding classification labels (107).

2. The method according to claim 1, characterized in that According to the preset classification rules, the corresponding classification labels (107) are respectively marked on the plurality of groups of fourth time series data, including: -classifying the plurality of sets of fourth time series data into at least two categories according to a preset classification rule; - Label the fourth time series data of the first category with the first sample; - Label the fourth time series data of the second category with the second sample.

3. The method according to claim 1, characterized in that Obtaining (103) the control signal inside the circuit breaker includes: -Get the current signal of the opening and closing coil; -Get the signal to control the opening and closing of contacts.

4. The method according to claim 1, wherein The preset filter fusion algorithms include: Kalman filter algorithm or particle filter algorithm.

5. The method according to claim 1, wherein Converting (102) the time-series rotational angular displacement data of the main shaft into time-series linear displacement data of the contact includes: -By calculating the four-bar linkage model in the circuit breaker, the model proportional coefficient k is obtained, and the rotation angle displacement data of the main shaft in time series is converted to Converted into linear displacement data f of the contact in time series, 6. The method according to claim 1, characterized in that Obtaining the first time series data includes: - obtain first time series data stored in array form; or, - Obtaining first time series data stored in a curve form.

7. A neural network model training method, characterized in that: include: - executing (301) the training data generation method according to any one of claims 1 to 6 to obtain a training data set containing a preset amount of training data; - Using the training data set (302) to train the neural network model to obtain an optimized neural network model; wherein the neural network model includes: a CNN model or an LSTM model.

8. An electronic device comprising a processor (401) and a memory (402), wherein the memory (402) stores computer-readable instructions, and when the computer-readable instructions are executed by the processor (401), the method according to any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed, perform the method according to any one of claims 1 to 7.

Images