Laser output power regulation and control method, device and equipment and storage medium
By precisely controlling the laser output power in the laser scribing process of perovskite solar cells, and utilizing the PID algorithm and the TCN-attention hybrid model, the problems of over-etching and material residue caused by laser power fluctuations were solved, thereby improving production stability and yield.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-04-03
AI Technical Summary
In the laser scribing process for perovskite solar cells, issues such as excessive etching or material residue caused by laser power fluctuations can affect production yield and stability.
By collecting multiple process parameters, the laser output power is precisely controlled using a PID algorithm and a temporal convolutional network model (TCN-attention hybrid model). Combined with sliding window analysis and gradient judgment, the half-wave plate angle and laser drive current are adjusted to achieve stable control of the laser output power.
This effectively reduced laser power fluctuations, improved the stability and production yield of laser scribing, and ensured efficient current collection of perovskite solar cells.
Smart Images
Figure CN121776684A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of laser output power control technology for laser marking, and in particular to a laser output power regulation method, device, equipment, and storage medium. Background Technology
[0002] In the industrial production of perovskite solar cells, laser scribing (P1, P2, P3 processes) is a key step in the fabrication of tandem cell structures. Its core purpose is to divide a large-area continuous thin film into multiple interconnected sub-cell units through high-precision laser etching, thereby achieving efficient current collection.
[0003] However, in the laser scribing process for perovskite solar cells, the thin film structure of perovskite solar cells is extremely fine (the film thickness of the coating material is usually in the nanometer to micrometer range, such as about 500 nm for the P1 layer and about 1-2 μm for the P2 layer), and the material properties of different functional layers (such as transparent conductive layer, perovskite light-absorbing layer, charge transport layer, etc.) are different. Therefore, different materials used for coating and different film thicknesses in the coating process require adjustment of the laser output power for process matching.
[0004] However, in actual production, lasers are susceptible to power fluctuations due to various factors. These include internal factors such as the instability of the internal discharge plasma, diode junction temperature drift, and power supply ripple, as well as external interferences such as the thermal lensing effect of the optical system lenses and the mechanical vibration of the galvanometer system. Power fluctuations are particularly sensitive in the processing of micron-scale perovskite thin films. Excessive power fluctuations, such as excessive power increases leading to over-etching and damage to the perovskite film material, can cause thermal damage and even short circuits. Conversely, excessive power drops can result in material residue causing series connection of conductive channels between sub-cells, forming current bypasses.
[0005] Since ensuring power stability during laser scribing is crucial for improving the production yield of perovskite solar cells, the laser scribing process has extremely stringent requirements for the stability of laser power. Summary of the Invention
[0006] The purpose of this invention is to provide a laser output power control method, device, equipment, and storage medium, which effectively reduces laser power fluctuations and improves the stability of laser scribing by precisely controlling the output power of the laser.
[0007] To achieve the above objectives, the present invention discloses a laser output power control method, comprising: Multiple process parameters are collected to form a process parameter set, the process parameters including laser power; Calculate the average power of the laser within a set sliding window; Determine whether the mean value is within the mean range. If it is not within the mean range, calculate the corrected output power value of the laser using a PID algorithm. Adjust the PID controller based on the calculated corrected output power value; Calculate the variance of the laser power within a set sliding window; Determine whether the variance is greater than a variance threshold; if it is greater than the variance threshold, then obtain the process parameter set. The parameters in the process parameter set are input into the trained power prediction model to obtain the predicted power set; Adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
[0008] Furthermore, the control methods also include: Calculate the gradient of the laser power within a set sliding window; The system determines whether the gradient is positive or negative. If it is positive, it determines whether the magnitude of the gradient change is greater than a first magnitude threshold. If it is greater than the first magnitude threshold, it reduces the driving current of the laser. If it is negative, it determines whether the magnitude of the gradient change is greater than a second magnitude threshold. If it is greater than the second magnitude threshold, it sends a heat dissipation detection signal.
[0009] Furthermore, the "calculation of the corrected output power value of the laser using a PID algorithm" includes: The corrected output power value of the laser is calculated using a first formula, which includes:
[0010] in The input signal is the energy output signal of the PID controller. For proportional adjustment, Kp is the proportional gain parameter, and e(t) is the error between the set value and the actual value of the laser output energy at the current moment; For integral term adjustment, Ki is the integral gain parameter. The integral accumulated for the error value at the current moment; For differential term adjustment, Kd is the differential gain parameter. This is the derivative of the error with respect to time.
[0011] Furthermore, the process parameters also include laser frequency, pump source current, pump source temperature, real-time coolant flow rate, and duty cycle.
[0012] Furthermore, before the phrase "collecting multiple process parameters to form a process parameter set, the process parameters including laser power" is mentioned, the method further includes: Collect multiple process parameters under normal and abnormal operating conditions to form a training parameter set; Perform a Fourier transform on the parameters in the training parameter set to obtain the first dataset; The data in the first dataset is filtered for interference sources to obtain the second dataset; Perform an inverse Fourier transform on the data in the second dataset to obtain the third dataset; Configure the power prediction model; The configured power prediction model is trained using data from the third dataset.
[0013] Furthermore, "configuring the power prediction model" includes: Define a temporal convolutional network module; At least five temporal convolutional network layers are constructed within the temporal convolutional network module; An attention mechanism is introduced to work in conjunction with the temporal convolutional network module to obtain the power prediction model.
[0014] Furthermore, the step of "performing interference source filtering on the data in the first dataset to obtain the second dataset" includes: The local maximum condition in the spectrum peak detection algorithm is used to perform spectrum peak detection on the data in the first dataset to identify the interference source frequency bands. The local maximum condition is as follows:
[0015] Where A[k] is the amplitude value at frequency k, and Tmin is the minimum amplitude threshold of the spectrum; The interference source frequency bands in the identified first dataset are filtered according to the second formula, which is:
[0016] The filtered data is then used as the data in the second dataset to obtain the second dataset.
[0017] To achieve the above objectives, the present invention discloses a laser output power control device, comprising: A collection module is used to collect multiple process parameters to form a process parameter set, the process parameters including laser power; The first calculation module is used to calculate the average value of the laser power within a set sliding window; The first judgment module is used to determine whether the mean value is within the mean value range. If it is not within the mean value range, the corrected output power value of the laser is calculated by the PID algorithm. The first adjustment module is used to adjust the PID controller based on the calculated corrected output power value; The second calculation module is used to calculate the variance of the laser power within a set sliding window; The second judgment module is used to determine whether the variance is greater than the variance threshold. If it is greater than the variance threshold, the process parameter set is obtained. The input module is used to input the parameters in the process parameter set into the trained power prediction model to obtain the predicted power set; The second adjustment module is used to adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
[0018] To achieve the above objectives, the present invention discloses an electronic device comprising: One or more processors; One or more memories are used to store one or more programs, which, when executed by the processor, cause the processor to implement the laser output power control method as described above.
[0019] To achieve the above objectives, the present invention discloses a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the laser output power control method as described above.
[0020] Compared with existing technologies, this invention analyzes and controls the output power of the laser by collecting multiple process parameters, including the laser power. First, it determines whether the laser power is within the mean range by calculating the mean value within a set sliding window. If not, it calculates the corrected output power value of the laser using a PID algorithm and adjusts the PID controller accordingly. Next, it determines whether the variance of the laser power within the set sliding window is greater than the variance threshold. If so, it obtains the process parameter set and inputs the parameters into the trained power prediction model to predict the power set, thereby adjusting the angle of the half-wave plate and / or the driving current of the laser. This effectively achieves precise control of the laser output power, effectively reduces laser power fluctuations, and improves the stability of laser scribing. Attached Figure Description
[0021] Figure 1 This is a flowchart of a laser output power control method according to an embodiment of the present invention.
[0022] Figure 2 This is a schematic diagram of the laser structure in the laser output power control method of this invention.
[0023] Figure 3 This is a schematic diagram of the power prediction model in the laser output power control method of this invention.
[0024] Figure 4This is a schematic diagram of the temporal convolutional network layer in the laser output power control method of this invention.
[0025] Figure 5 This is a schematic diagram of the attention mechanism in the laser output power control method of this invention.
[0026] Figure 6 This is a block diagram of a laser output power control device according to an embodiment of the present invention.
[0027] Figure 7 This is a system diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation
[0028] To illustrate the technical content, structural features, objectives, and effects of the present invention in detail, the following description is provided in conjunction with the embodiments and accompanying drawings.
[0029] Example 1 Please see Figures 1 to 5 This invention discloses a method for controlling the output power of a laser, comprising: It should be noted that the control method is used to address, for example, Figure 2 The output power of the laser shown is regulated. The laser is equipped with a high-precision half-wave plate, a polarizing beam splitter (PBS), and a high-sensitivity power meter mounted on a motor. The polarizing beam splitter splits a beam of light into two mutually perpendicular beams: a main processing beam (approximately 90% energy) and a monitoring beam (approximately 10% energy). The energy fluctuation of the monitoring beam is measured in real time by a high-precision, low-delay, and highly sensitive photodiode-type power meter. The power meter sensor, designed based on the photoelectric effect of a semiconductor PN junction, can achieve a nanosecond-level response speed, meeting the requirements for dynamic power adjustment. The host computer control system processes and analyzes the signal fed back from the power meter, and adjusts the output power of the laser through a PID controller or adjusts the rotation angle of the half-wave plate of the optical device through the motor to dynamically adjust the polarization state of the incident laser, thereby forming a closed-loop regulation and control system for monitoring and predicting the fluctuation of the laser's output power and regulating the power of the main processing beam.
[0030] Furthermore, before "collecting multiple process parameters to form a process parameter set, including laser power," it also includes: 11. Collect multiple process parameters under normal and abnormal operating conditions to form a training parameter set; Furthermore, multiple process parameters include laser power, laser frequency, pump source current, pump source temperature, real-time coolant flow rate, and duty cycle.
[0031] Collecting data on various process and operating parameters that may affect laser output power under both normal and abnormal operating conditions during the production process to form a model training dataset is beneficial for improving the accuracy of model predictions.
[0032] 12. Perform Fourier transform on the parameters in the training parameter set to obtain the first dataset; Among the various process and operating parameters collected, interference information is inevitable. For example, the current and voltage of the laser pump source may fluctuate due to the power supply voltage. White noise such as dust obstruction and galvanometer vibration in the environment will also affect the power of the laser. Frequency domain analysis can identify interference sources such as voltage and current fluctuations, galvanometer vibration, and half-wave plate vibration, which facilitates subsequent noise reduction.
[0033] It is understood that in this embodiment, multiple process parameters such as laser power, laser frequency, pump source current, pump source temperature, real-time coolant flow rate, and duty cycle are processed using Discrete Fourier Transform (DFT) to extract frequency domain features. The specific DFT formula is as follows:
[0034] Where K is an integer index representing the component number in the frequency domain, which physically represents the identifier of different frequency components in the input signal; n is another integer index representing the sampling point position of the time domain signal, which physically represents the specific sampling point of the input signal in the time sequence; X[k] is the frequency domain complex sequence obtained after calculation (containing amplitude / phase information); x[n] is the time sequence sampling data of multiple process parameters (including laser power, laser frequency, pump source current, pump source temperature, real-time flow rate of coolant, duty cycle, etc.); N is the number of sampling points, and j is the imaginary unit.
[0035] 13. Perform interference source filtering on the data in the first dataset to obtain the second dataset; Identifying interference sources and performing noise reduction filtering on the frequency domain data after DFT transformation is beneficial for optimizing long-term process performance.
[0036] Furthermore, "performing interference source filtering on the data in the first dataset to obtain the second dataset" includes: 131. Utilize the local maximum condition in the spectrum peak detection algorithm to perform spectrum peak detection on the data in the first dataset to identify the interference source frequency bands. The local maximum condition is:
[0037] Where A[k] is the amplitude value at frequency k, and Tmin is the minimum amplitude threshold of the spectrum; 132. Based on the second formula, the interference source frequency bands in the identified first dataset are filtered. The second formula is:
[0038] 133. The filtered data becomes part of the second dataset, thus obtaining the second dataset.
[0039] 14. Perform an inverse Fourier transform on the data in the second dataset to obtain the third dataset; It is understood that in this embodiment, the time-domain data is reconstructed through the Inverse Fourier Transform (DFT) as input data, and the denoised time-domain signal after the transformation is used as the input data for the subsequent TCN-attention hybrid prediction model for training. The specific DFT formula is as follows:
[0040] 15. Configure the power prediction model; Furthermore, the "configuration power prediction model" includes: 151. Define a temporal convolutional network module; It is understood that in this embodiment, a temporal convolutional network (TCN) is used as the core architecture. Its dilated causal convolutional structure can effectively capture the long-term and short-term dependencies of power fluctuations. Furthermore, various parameters in the temporal convolutional network module are set, such as selecting the activation function, whether to use padding, whether to use dilation, etc., to realize the definition of the temporal convolutional network module.
[0041] 152. Construct at least five temporal convolutional network layers within the temporal convolutional network module; It is understandable that 'd' in a temporal convolutional network represents the dilation factor of each layer. The exponential increase of the dilation factor in each layer expands the receptive field. In this embodiment, the temporal convolutional network (TCN) needs at least 5 layers to sequentially expand the dilation factor, effectively capturing the fluctuation characteristics of laser power at multiple scales. Furthermore, the output and input of each layer are connected through residuals to ensure a stable gradient propagation model and construct a structure as shown below. Figure 4 The temporal convolutional network layer shown is used for convolution operations along the sampling time series dimension of the process parameters to combine the corresponding input process, operating condition features and time dimension, thereby extracting features on the time axis and predicting future power fluctuations based on the temporal data of the input time window.
[0042] 153. An attention mechanism is introduced to work in conjunction with a temporal convolutional network module to obtain a power prediction model.
[0043] It is understood that, in this embodiment, the following will be used: Figure 5The attention mechanism shown introduces a power prediction model to dynamically weight and focus on the fluctuation characteristics of key time points (e.g., the first 100ms of a power surge).
[0044] 16. Use the data in the third dataset to train the configured power prediction model.
[0045] It is understood that, in this embodiment, the architecture constructed using the multi-dimensional feature vectors built in steps 11 to 14 above is as follows: Figure 3 The TCN-attention hybrid power prediction model shown is trained to model temporal features.
[0046] 101. Collect multiple process parameters to form a process parameter set, including laser power; Furthermore, several process parameters also include laser frequency, pump source current, pump source temperature, real-time coolant flow rate, and duty cycle.
[0047] 102. Calculate the average laser power within a set sliding window; Understandably, multi-scale sliding window technology is used to extract the time-domain characteristics of power fluctuations for time-domain analysis in order to perform corresponding control operations. This involves calculating the mean, variance, and gradient of the laser output power over a certain time range. The formula for the sliding window is as follows: , where n is the window length.
[0048] It should be noted that the formula for calculating the average power over a certain time range using a sliding window is as follows:
[0049] 103. Determine whether the mean is within the mean range. If it is not within the mean range, calculate the corrected output power value of the laser using the PID algorithm. It should be noted that the average range is an open and settable range. The judgment and operation standards for power fluctuations need to be formulated according to the laser's maintenance manual or actual conditions. For example, if a 20W laser is selected, and the splitting ratio is adjusted to 1:10, and the deviation and fluctuation of the power meter are added, the average range can be set to 0.9W-1.1W. The calculated average can reflect the baseline level of power and be used to determine whether to trigger the integral term adjustment of the PID controller.
[0050] Furthermore, "calculating the corrected output power value of the laser using a PID algorithm" includes: 1031. Calculate the corrected output power value of the laser using the first formula, which includes:
[0051] in The input signal is the energy output signal of the PID controller. For proportional adjustment, Kp is the proportional gain parameter, and e(t) is the error between the set value and the actual value of the laser output energy at the current moment; For integral term adjustment, Ki is the integral gain parameter, which represents the strength of the cumulative error's influence on the controller output. The integral accumulated over the current error value represents the total accumulated amount of historical error. For differential term adjustment, Kd is the differential gain parameter. The derivative of the error with respect to time represents the rate at which the error changes at the current moment.
[0052] 104. Adjust the PID controller based on the calculated corrected output power value; By fine-tuning the PID controller based on real-time power meter data and by calculating the corrected output power value (adjustment amount) of the laser through historical cumulative deviation, residual steady-state errors can be effectively eliminated.
[0053] 105. Calculate the variance of the laser power within a set sliding window; It is understood that, in this embodiment, the formula for calculating the power variance over a certain time range using a sliding window is as follows:
[0054] 106. Determine if the variance is greater than the variance threshold. If it is greater than the variance threshold, obtain the process parameter set. 107. Input the parameters from the process parameter set into the trained power prediction model to obtain the predicted power set; 108. Adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
[0055] It should be noted that the variance threshold is an open and settable threshold. The judgment and operation standards for power fluctuation abrupt changes need to be formulated according to the laser's maintenance manual. The calculated variance can quantify the severity of power fluctuations to identify abnormal operating conditions such as voltage fluctuation interference or abnormal light source components, in order to determine whether it is necessary to start the TCN model prediction to obtain the TCN prediction results of the power value over a certain time span in the future. And by adjusting the angle of the half-wave plate and / or the laser's drive current in advance (requiring the laser's response time to be at least in milliseconds), possible step fluctuations can be compensated in advance.
[0056] The purpose of the aforementioned real-time time-domain analysis and control is to quickly detect sudden changes in order to maintain short-term power stability. Specifically, based on the step fluctuation results predicted by the TCN, the half-wave plate angle or laser drive current is adjusted in advance, with a control cycle of up to 10ms, meeting the real-time requirements of industry and achieving feedforward control. At the same time, the PID controller is adjusted according to the corrected output power value to achieve feedback control. The two work together to form a feedforward-feedback composite control to precisely regulate the laser output power. For large-format glass plates (not limited to other materials), through the aforementioned steps of acquiring laser process data, operating condition auxiliary data, optimizing neural networks, training models, predicting power fluctuations, and implementing feedforward-feedback composite control, the power fluctuations of the laser can be effectively reduced, thereby achieving stable regulation of the laser output power for perovskite laser scribing and stabilizing the control of the thermal effect zone of perovskite laser scribing.
[0057] Furthermore, the control methods also include: 109. Calculate the gradient of laser power within a set sliding window; It is understandable that the power gradient over a certain time range can be obtained by calculating the laser power data points (i, pi) using the following formula:
[0058] Where i is the time index; The calculated gradient can capture the trend and rate of power change, so as to accurately distinguish and identify slow temperature drift (such as thermal lensing effect) and sudden power fluctuations.
[0059] 110. Determine whether the gradient is positive or negative. If it is positive, determine whether the gradient change amplitude is greater than the first amplitude threshold. If it is greater than the first amplitude threshold, reduce the laser's drive current. If it is negative, determine whether the gradient change amplitude is greater than the second amplitude threshold. If it is greater than the second amplitude threshold, issue a heat dissipation detection signal.
[0060] It should be noted that in this embodiment, the first amplitude threshold and the second amplitude threshold are open and settable thresholds. The judgment and operation standards for power fluctuation changes need to be formulated according to the laser's maintenance manual. When the gradient of the laser output power changes abruptly, it is usually necessary to check whether the hardware is faulty. When the change amplitude of the gradient is greater than the first amplitude threshold, it indicates a sudden increase in the positive gradient, which may be due to plasma instability. In this case, the laser current needs to be reduced. When the change amplitude of the gradient is greater than the second amplitude threshold, it indicates a sudden drop in the negative gradient, which may be due to a failure in the cooling system. In this case, the heat dissipation module needs to be checked.
[0061] Compared with existing technologies, this invention analyzes and controls the output power of the laser by collecting multiple process parameters, including the laser power. First, it determines whether the laser power is within the mean range by calculating the mean value within a set sliding window. If not, it calculates the corrected output power value of the laser using a PID algorithm and adjusts the PID controller accordingly. Next, it determines whether the variance of the laser power within the set sliding window is greater than the variance threshold. If so, it obtains the process parameter set and inputs the parameters into the trained power prediction model to predict the power set, thereby adjusting the angle of the half-wave plate and / or the driving current of the laser. This effectively achieves precise control of the laser output power, effectively reduces laser power fluctuations, and improves the stability of laser scribing.
[0062] Example 2 Please refer to the figure. Figures 1 to 6 This invention discloses a laser output power control device, which includes: The collection module 201 is used to collect multiple process parameters to form a process parameter set, including laser power; The first calculation module 202 is used to calculate the average value of the laser power within a set sliding window; The first judgment module 203 is used to determine whether the mean is within the mean range. If it is not within the mean range, the corrected output power value of the laser is calculated by the PID algorithm. The first adjustment module 204 is used to adjust the PID controller according to the calculated corrected output power value; The second calculation module 205 is used to calculate the variance of the laser power within a set sliding window; The second judgment module 206 is used to determine whether the variance is greater than the variance threshold. If it is greater than the variance threshold, the process parameter set is obtained. Input module 207 is used to input the parameters in the process parameter set into the trained power prediction model to obtain the predicted power set; The second adjustment module 208 is used to adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
[0063] Example 3 Please refer to the figure. Figures 1 to 7 This invention discloses an electronic device comprising: One or more processors 301; One or more memories 302 are used to store one or more programs, which, when executed by a processor, enable the processor to implement the laser output power control method described above.
[0064] Example 4 This application discloses a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements the laser output power control method as described above.
[0065] Example 5 This application discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned laser output power control method.
[0066] It should be understood that, in the embodiments of this application, the processor may be a central processing unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.
[0067] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by hardware related to computer program instructions. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0068] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.
Claims
1. A method for controlling the output power of a laser, characterized in that, include: Multiple process parameters are collected to form a process parameter set, the process parameters including laser power; Calculate the average power of the laser within a set sliding window; Determine whether the mean value is within the mean range. If it is not within the mean range, calculate the corrected output power value of the laser using a PID algorithm. Adjust the PID controller based on the calculated corrected output power value; Calculate the variance of the laser power within a set sliding window; Determine whether the variance is greater than a variance threshold; if it is greater than the variance threshold, then obtain the process parameter set. The parameters in the process parameter set are input into the trained power prediction model to obtain the predicted power set; Adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
2. The laser output power control method according to claim 1, characterized in that, Also includes: Calculate the gradient of the laser power within a set sliding window; The system determines whether the gradient is positive or negative. If it is positive, it determines whether the magnitude of the gradient change is greater than a first magnitude threshold. If it is greater than the first magnitude threshold, it reduces the driving current of the laser. If it is negative, it determines whether the magnitude of the gradient change is greater than a second magnitude threshold. If it is greater than the second magnitude threshold, it sends a heat dissipation detection signal.
3. The laser output power control method according to claim 1, characterized in that, The "calculation of the corrected output power value of the laser using a PID algorithm" includes: The corrected output power value of the laser is calculated using a first formula, which includes: in The input signal is the energy output signal of the PID controller. For proportional adjustment, Kp is the proportional gain parameter, and e(t) is the error between the set value and the actual value of the laser output energy at the current moment; For integral term adjustment, Ki is the integral gain parameter. The integral accumulated for the error value at the current moment; For differential term adjustment, Kd is the differential gain parameter. This is the derivative of the error with respect to time.
4. The laser output power control method according to claim 1, characterized in that, The process parameters also include laser frequency, pump source current, pump source temperature, real-time coolant flow rate, and duty cycle.
5. The laser output power control method according to claim 1, characterized in that, Before the phrase "collecting multiple process parameters to form a process parameter set, the process parameters including laser power", the following is also included: Collect multiple process parameters under normal and abnormal operating conditions to form a training parameter set; Perform a Fourier transform on the parameters in the training parameter set to obtain the first dataset; The data in the first dataset is filtered for interference sources to obtain the second dataset; Perform an inverse Fourier transform on the data in the second dataset to obtain the third dataset; Configure the power prediction model; The configured power prediction model is trained using data from the third dataset.
6. The laser output power control method according to claim 5, characterized in that, The "configuring the power prediction model" includes: Define a temporal convolutional network module; At least five temporal convolutional network layers are constructed within the temporal convolutional network module; An attention mechanism is introduced to work in conjunction with the temporal convolutional network module to obtain the power prediction model.
7. The laser output power control method according to claim 5, characterized in that, The phrase "performing interference source filtering on the data in the first dataset to obtain the second dataset" includes: The local maximum condition in the spectrum peak detection algorithm is used to perform spectrum peak detection on the data in the first dataset to identify the interference source frequency bands. The local maximum condition is as follows: Where A[k] is the amplitude value at frequency k, and Tmin is the minimum amplitude threshold of the spectrum; The interference source frequency bands in the identified first dataset are filtered according to the second formula, which is: The filtered data is then used as the data in the second dataset to obtain the second dataset.
8. A laser output power control device, characterized in that, include: A collection module is used to collect multiple process parameters to form a process parameter set, the process parameters including laser power; The first calculation module is used to calculate the average value of the laser power within a set sliding window; The first judgment module is used to determine whether the mean value is within the mean value range. If it is not within the mean value range, the corrected output power value of the laser is calculated by the PID algorithm. The first adjustment module is used to adjust the PID controller based on the calculated corrected output power value; The second calculation module is used to calculate the variance of the laser power within a set sliding window; The second judgment module is used to determine whether the variance is greater than the variance threshold. If it is greater than the variance threshold, the process parameter set is obtained. The input module is used to input the parameters in the process parameter set into the trained power prediction model to obtain the predicted power set; The second adjustment module is used to adjust the angle of the half-wave plate and / or the drive current of the laser based on the predicted power set.
9. An electronic device, characterized in that, include: One or more processors; One or more memories for storing one or more programs, which, when executed by the processor, cause the processor to implement the laser output power control method as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the laser output power control method as described in any one of claims 1 to 7.