Circuit operation parameter adjusting method and device

Abstract

The invention discloses a method and a device for adjusting circuit operation parameters. The method comprises the following steps: acquiring initial circuit operation parameters of equipment heated by using plasma; inputting the initial circuit operation parameters into a preset first model for plasma stability and system energy consumption calculation to obtain an initial stability value and an initial system energy consumption value of the circuit; taking the initial circuit operation parameter as an initial position parameter of each particle, and performing fitness calculation according to the initial position parameter, the initial stability value, the initial system energy consumption value, the first model and a preset particle swarm optimization algorithm to obtain a target circuit operation parameter which enables the plasma stability value to be maximum and the system energy consumption value to be minimum; and adjusting the initial circuit operation parameter to the target circuit operation parameter. According to the scheme, the working parameters of the booster circuit are dynamically adjusted through the preset particle swarm optimization algorithm, the stability of the plasma can be remarkably improved, the energy consumption is reduced, and the working efficiency of equipment using the plasma for heating is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit optimization, and in particular to a method and device for adjusting circuit operating parameters. Background Art

[0002] As an important industrial equipment, the working efficiency of the booster circuit and the stability of the plasma of the equipment using plasma heating directly affect the performance of the equipment. In the prior art, the booster circuit efficiency and plasma stability of the equipment using plasma heating usually rely on empirical adjustment and fixed parameter settings. This method cannot be adaptively optimized according to the real-time working status, resulting in high energy consumption and poor system stability. Summary of the invention

[0003] The present invention aims to at least solve the technical problems existing in the prior art. To this end, the present invention provides a circuit operation parameter adjustment method in a first aspect, the method comprising: Acquiring initial circuit operating parameters of the device using plasma heating in a working state, the initial circuit operating parameters including: an input voltage value of a boost circuit, a current value in a circuit, an operating frequency of the circuit, and an electrode spacing; the electrode spacing represents the spacing between a discharge electrode and a ground electrode of a plasma generator of the device using plasma heating; the device using plasma heating is a device that converts electrical energy into a high-temperature plasma flame using a plasma heating principle; Inputting the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, and obtaining an initial stability value and an initial system energy consumption value of the circuit; The initial circuit operation parameters are used as initial position parameters of each particle, and fitness calculation is performed according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm to obtain target circuit operation parameters that maximize the plasma stability value and minimize the system energy consumption value; The initial circuit operating parameters are adjusted to the target circuit operating parameters.

[0004] Optionally, the first model includes a feature extraction layer, a feature fusion layer, a self-attention mechanism layer and a fully connected layer, and the initial circuit operation parameters are input into a preset first model to calculate plasma stability and system energy consumption to obtain an initial stability value and an initial system energy consumption value of the circuit, including: Inputting the initial circuit operation parameters into the feature extraction layer and the feature fusion layer to perform initial feature extraction and feature fusion to obtain a first feature value; Input the first eigenvalue into the self-attention mechanism layer, adaptively calculate the target feature weight corresponding to each feature in the first eigenvalue by combining maximum pooling and average pooling, and weight the first eigenvalue by the target feature weight to obtain a second eigenvalue; Inputting the second eigenvalue into the feature extraction layer for further feature extraction to obtain a third eigenvalue; The third eigenvalue is input into a fully connected layer for full connection processing, and an initial stability value and an initial system energy consumption value of the circuit are output.

[0005] Optionally, the feature extraction layer includes a first convolution layer, a second convolution layer, a third convolution layer, a fourth convolution layer and a fifth convolution layer, and the initial circuit operation parameter is input into the feature extraction layer and the feature fusion layer to perform initial feature extraction and fusion to obtain a first feature value, including: Inputting the first eigenvalue into the first convolutional layer for feature extraction to obtain a first sub-eigenvalue; Inputting the first eigenvalue into the second convolutional layer for feature extraction to obtain a second sub-eigenvalue; Inputting the second eigenvalue into the third convolutional layer for feature extraction to obtain a third sub-eigenvalue; Concatenate the first sub-eigenvalue and the third sub-eigenvalue to obtain a fourth sub-eigenvalue; Inputting the fourth sub-eigenvalue into the fourth convolutional layer for feature extraction to obtain a fifth sub-eigenvalue; The second sub-eigenvalue and the fifth sub-eigenvalue are concatenated to obtain a first eigenvalue.

[0006] Optionally, the loss function of the first model during training includes a first error term, a second error term and a regularization term; The first error term represents the mean square error between the predicted plasma stability value and the actual plasma stability value; the second error term represents the mean square error between the predicted system energy consumption value and the actual system energy consumption value; the regularization term is used to control the physical rationality of the circuit operating parameters.

[0007] Optionally, the using the initial circuit operation parameters as initial position parameters of each particle, and performing fitness calculation according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm, includes: Initialize particle swarm parameters; the particle swarm parameters include at least the number of particles, the position range and speed range of each particle, the inertia weight of the particle, the initial stability weight and the initial energy consumption weight; the position range is from the minimum value to the maximum value of the circuit operation parameter; Using the initial circuit operation parameters as initial position parameters of each particle; Randomly generate an initial position and an initial speed of each particle in an initial state within the position range and the speed range, and generate an initial individual optimal position and an initial global optimal position of each particle; Calculating the initial fitness of each particle according to the initial stability weight, the initial energy consumption weight, the initial position parameter, the initial system energy consumption value, the initial stability value and a preset fitness function; According to the particle group parameters and a preset speed update function, the speed of the particles is updated within the speed range to obtain an updated speed; According to the update speed and the initial position of the particle, updating the position of the particle within the position range to obtain an updated position; Calculating a new fitness of the particle based on the position parameter of the particle at the updated position; If the new fitness is greater than the individual historical optimal value, the new fitness value is used as the current individual optimal value; if the new fitness is greater than the global historical optimal value, the new fitness value is used as the current global optimal value; When the preset maximum number of iterations is reached or the fitness reaches a preset convergence threshold, the calculation is stopped, the position parameters of the particles located at the global optimal value are obtained, and the optimal circuit parameters are obtained; the optimal circuit parameters are the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value.

[0008] Optionally, updating the particle's speed within the speed range according to the initial parameters and a preset speed update function to obtain an updated speed includes: Calculating a first difference between the initial individual optimal position and the initial position of the particle, and a second difference between the initial global optimal position and the initial position; Determine the product of the first difference, the individual learning factor, and a preset first random number to obtain a first product; and determine the product of the second difference, the global learning factor, and a preset second random number to obtain a second product; determining the product of the inertia weight and the initial velocity of the particle to obtain a third product; Determine the sum of the first product, the second product, and the third product to obtain an initial update speed; If the initial update speed is within the speed range, the initial update speed is used as the update speed.

[0009] Optionally, the calculating the initial fitness of each particle according to the initial stability weight, the initial energy consumption weight, the initial position parameter, the initial system energy consumption value, the initial stability value and a preset fitness function includes: Inputting the initial position parameters into the first model to obtain an initial plasma stability value and an initial system energy consumption value; Substituting the initial position parameter into a preset regularization term to obtain a regularization value; Determine the product of the initial stability weight and the initial plasma stability value to obtain a stability product; and determine the product of the initial energy consumption weight and the initial system energy consumption value to obtain an energy consumption product; The difference between the energy consumption product and the stability product is determined, and the sum of the difference and the regularization value is determined, and the sum is used as the initial fitness of the particle.

[0010] A second aspect of the present invention provides a circuit operation parameter adjustment device, the device comprising: A parameter acquisition module, used to acquire initial circuit operation parameters of the device using plasma heating when it is in working state, wherein the initial circuit operation parameters include: input voltage value of the boost circuit, current value in the circuit, operating frequency of the circuit and electrode spacing; the electrode spacing represents the spacing between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating; the device using plasma heating refers to a device using the plasma heating principle as a heat source; A calculation module, used for inputting the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, so as to obtain an initial stability value and an initial system energy consumption value of the circuit; a target parameter determination module, configured to use the initial circuit operation parameters as the initial position parameters of each particle, perform fitness calculation according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm, and obtain the target circuit operation parameters that maximize the plasma stability value and minimize the system energy consumption value; The adjustment module is used to adjust the initial circuit operation parameters to the target circuit operation parameters.

[0011] The third aspect of the present invention provides a device using plasma heating, comprising: a controller; a memory for storing executable instructions of the controller; wherein the controller is configured to execute the instructions to implement the circuit operation parameter adjustment method as described in the first aspect.

[0012] A fourth aspect of the present invention provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a controller of a device using plasma heating, the device using plasma heating can perform the circuit operation parameter adjustment method described in the first aspect.

[0013] The embodiments of the present invention have the following beneficial effects: In an embodiment of the present invention, the initial circuit operating parameters of the device using plasma heating in the working state are obtained, and the initial circuit operating parameters include: the input voltage value of the boost circuit, the current value in the circuit, the operating frequency of the circuit and the electrode spacing; the electrode spacing represents the spacing between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating; the device using plasma heating refers to a device using the plasma heating principle as a heat source; the initial circuit operating parameters are input into a preset first model to calculate the plasma stability and system energy consumption, and the initial stability value and initial system energy consumption value of the circuit are obtained; the initial circuit operating parameters are used as the initial position parameters of each particle, and the fitness is calculated according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and the preset particle swarm optimization algorithm to obtain the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value; the initial circuit operating parameters are adjusted to the target circuit operating parameters. This scheme dynamically adjusts the working parameters of the boost circuit through the preset particle swarm optimization algorithm, which can significantly improve the stability of the plasma, reduce energy consumption, and improve the working efficiency of the device using plasma heating. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flow chart of the steps of a method for adjusting circuit operation parameters provided by an embodiment of the present invention; Figure 2 It is a structural block diagram of a circuit operation parameter adjustment device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, unless otherwise specified, "multiple" means two or more. In addition, the use of "based on" or "according to" means openness and inclusiveness, because the process, steps, calculations or other actions "based on" or "according to" one or more of the conditions or values ​​may be based on additional conditions or values ​​beyond the described values ​​in practice.

[0017] Figure 1 It is a flow chart of the steps of a method for adjusting circuit operation parameters provided by an embodiment of the present invention.

[0018] like Figure 1 As shown, the method comprises the following steps: Step 101, obtaining initial circuit operating parameters of the device using plasma heating when it is in working state, wherein the initial circuit operating parameters include: an input voltage value of a boost circuit, a current value in the circuit, an operating frequency of the circuit and an electrode spacing; the electrode spacing represents the spacing between a discharge electrode and a ground electrode of a plasma generator of the device using plasma heating; the device using plasma heating refers to a device using a plasma heating principle as a heat source.

[0019] In the embodiments of the invention, the device using plasma heating refers to a device using the plasma heating principle as a heat source, and the device includes but is not limited to a stove, an electric heater, a wall-mounted boiler, an electric oven, a portable stove, etc. The plasma heating principle refers to generating an open flame by breaking through the air with high pressure at the end of the plasma, and heating the related equipment by the generated open flame.

[0020] The device using plasma heating includes a boost circuit for boosting the mains voltage. The circuit operation parameters of the device using plasma heating in the working state are obtained, and the circuit operation parameters include: the input voltage value (V) of the boost circuit, the current value (I) in the circuit, the operating frequency (f) of the circuit, and the electrode spacing (d).

[0021] Here, the electrode distance refers to the distance between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating.

[0022] A data acquisition module can be deployed on the device using plasma heating, and the data acquisition module obtains the initial circuit operation parameters of the device using plasma heating when it is in working state. The data acquisition module sends the initial circuit operation parameters to the controller, and the controller executes the circuit operation parameter adjustment method of the present invention.

[0023] The controller can be deployed on a device using plasma heating or on the cloud, and technicians in this field can set it up according to actual application requirements.

[0024] Step 102: Input the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, and obtain an initial stability value and an initial system energy consumption value of the circuit.

[0025] The first model is a circuit simulation model built based on a deep neural network (DNN), which is used to determine the impact of circuit operating parameters on plasma stability and system energy consumption.

[0026] The input data of the first model is the circuit operating parameters. The features of the circuit operating parameters are extracted and processed through a deep neural network, and the estimated values ​​of plasma stability and system energy consumption are output.

[0027] Step 103: Use the initial circuit operating parameters as the initial position parameters of each particle, and perform fitness calculation based on the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm to obtain the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value.

[0028] The particle swarm optimization algorithm is to initialize a group of particles in the problem space. The particles move randomly through a specific algorithm and evaluate their movement quality through the fitness function. The particle swarm generates random solutions through the initialization phase and finds the optimal solution through iteration. In each iteration, the particles update themselves by tracking their own experience and group experience. After the update, the fitness value is calculated and the global optimal position and the local optimal position are updated. When the termination condition is judged to be met, a loop is proposed, otherwise the iteration continues. Finally, the optimal position and the optimal fitness value are output.

[0029] This scheme improves the particle swarm optimization (PSO) algorithm and designs an adaptive multi-objective particle swarm optimization algorithm to obtain the optimal circuit operating parameters.

[0030] The multi-objective particle swarm optimization algorithm introduces a multi-objective optimization method based on the standard PSO in order to simultaneously optimize the stability and energy consumption of the plasma beam. This method allows each particle to not only focus on a single objective, but also consider the balance of multiple objectives.

[0031] The algorithm adopts a weighted dynamic adjustment mechanism of the optimization target, a real-time feedback mechanism and an adaptive particle update strategy, and adds soft constraint guidance to improve the particle swarm optimization algorithm.

[0032] Among them, the weighted dynamic adjustment mechanism of the optimization target: the weights of stability (S) and energy consumption (E) in the PSO fitness function can be dynamically adjusted based on the real-time measurement of the plasma beam behavior. In the initial stage, more attention is paid to stability, and after the circuit has been running for a period of time, it turns to optimizing energy consumption.

[0033] Real-time feedback mechanism: When the circuit parameters change, the response of the plasma beam (such as fluctuations, stability changes, etc.) will be monitored in real time and fed back to the PSO algorithm. This can be done through multiple simulations or experimental data sets to obtain the real-time behavior of the plasma beam, and use this feedback information for the dynamic update of the particles.

[0034] Adaptive particle update strategy: Dynamically adjust the particle speed update method according to the stability and response of the plasma beam. For example, during the process of adjusting circuit parameters, the stability change of the plasma beam is evaluated in real time. If the stability of the plasma beam improves, the particles can be updated more significantly to explore more possible circuit configurations; if the stability decreases, the particle update amplitude is reduced to avoid over-adjustment. In the early stage of the search, a higher weight is given to the individual learning factor to promote extensive exploration; in the convergence stage, the group learning factor weight is gradually increased to help the algorithm find the optimal solution.

[0035] Add soft constraint guidance: By penalizing soft constraints such as excessive voltage, current, frequency and electrode spacing, it is effectively ensured that the optimized circuit parameters not only meet the performance requirements but also do not lead to unrealistic or unsafe working conditions, thereby enhancing the feasibility and stability of the algorithm.

[0036] This scheme uses the initial circuit operating parameters as the initial position parameters of each particle, and finds the optimal target circuit operating parameters by searching the space of voltage, current, frequency and electrode spacing to minimize the fitness function. When the fitness function is minimized, the plasma stability value is maximized and the system energy consumption value is minimized.

[0037] Step 104: Adjust the initial circuit operating parameters to the target circuit operating parameters.

[0038] Since the target circuit operating parameters can maximize the plasma stability value and minimize the system energy consumption value, adjusting the initial circuit operating parameters to the target circuit operating parameters can improve the boost circuit efficiency and plasma beam stability of the equipment using plasma heating.

[0039] In summary, in an embodiment of the present invention, the initial circuit operating parameters of the device using plasma heating in the working state are obtained, and the initial circuit operating parameters include: the input voltage value of the boost circuit, the current value in the circuit, the operating frequency of the circuit and the electrode spacing; the electrode spacing represents the spacing between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating; the device using plasma heating refers to a device using the plasma heating principle as a heat source; the initial circuit operating parameters are input into the preset first model to calculate the plasma stability and system energy consumption, and the initial stability value and initial system energy consumption value of the circuit are obtained; the initial circuit operating parameters are used as the initial position parameters of each particle, and the fitness is calculated according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and the preset particle swarm optimization algorithm to obtain the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value; the initial circuit operating parameters are adjusted to the target circuit operating parameters. This scheme dynamically adjusts the working parameters of the boost circuit through the preset particle swarm optimization algorithm, which can significantly improve the stability of the plasma, reduce energy consumption, and improve the working efficiency of the device using plasma heating.

[0040] As an optional embodiment, the first model includes a feature extraction layer, a feature fusion layer, a self-attention mechanism layer and a fully connected layer, and step 102 includes: Step 1021: input the initial circuit operation parameters into the feature extraction layer and the feature fusion layer to perform initial feature extraction and feature fusion to obtain a first feature value; Step 1022: input the first eigenvalue into the self-attention mechanism layer, adaptively calculate the target feature weight corresponding to each feature in the first eigenvalue by combining maximum pooling and average pooling, and weight the first eigenvalue by the target feature weight to obtain a second eigenvalue; Step 1023: input the second eigenvalue into a feature extraction layer for further feature extraction to obtain a third eigenvalue; Step 1024: input the third eigenvalue into a fully connected layer for full connection processing, and output an initial stability value and an initial system energy consumption value of the circuit.

[0041] In step 1021 to step 1024, feature extraction and feature fusion are performed on the initial circuit operation parameters to obtain a first feature value.

[0042] The feature extraction layer includes five convolutional layers, and a self-attention mechanism layer is introduced between the fourth and fifth convolutional layers to adjust the weight of the first feature value. The self-attention mechanism layer can capture the global dependency between input features and learn the importance of each feature.

[0043] Specifically, the maximum pooling algorithm and the average pooling algorithm are combined to adaptively calculate the weight of each layer of features in the first eigenvalue to obtain the target feature weight of each feature. The first eigenvalue is weighted and calculated using the target feature weight to obtain the second eigenvalue.

[0044] The second eigenvalue is input into the fifth convolutional layer of the feature extraction layer to perform feature extraction again to obtain the third eigenvalue.

[0045] After obtaining the third eigenvalue, the fully connected layer is used to further process the features. The fully connected layer includes the first fully connected layer and the second fully connected layer. The output dimension of the first fully connected layer is 64 neurons, and the output shape is [1, 64]; the output dimension of the second fully connected layer is 32 neurons, and the output shape is [1, 32].

[0046] The third eigenvalue is fully connected through the first fully connected layer and the second fully connected layer in turn, and an initial stability value and an initial system energy consumption value of the circuit are output.

[0047] The activation function used by the first model can be GELU (Gaussian Error Linear Unit). GELU can provide more refined gradient optimization, can finely adjust the relationship between the circuit and the plasma beam stability, and can improve the circuit efficiency and the accuracy of beam stability prediction.

[0048] The first model uses a self-attention mechanism to dynamically adjust the weight of each feature in the final output, helping the model understand the relationship between different circuit parameters and thus better predict the stability and energy consumption of the plasma. As an optional embodiment, the feature extraction layer includes a first convolution layer, a second convolution layer, a third convolution layer, a fourth convolution layer and a fifth convolution layer, and step 1021 includes: Step 10211: input the first eigenvalue into the first convolutional layer for feature extraction to obtain a first sub-eigenvalue; Step 10212: input the first eigenvalue into the second convolutional layer for feature extraction to obtain a second sub-eigenvalue; Step 10213: input the second eigenvalue into the third convolutional layer for feature extraction to obtain a third sub-eigenvalue; Step 10214: concatenate the first eigenvalue and the third eigenvalue to obtain a fourth sub-eigenvalue; Step 10215: input the fourth eigenvalue into the fourth convolutional layer for feature extraction to obtain a fifth sub-eigenvalue; Step 10216: Concatenate the second sub-eigenvalue and the fifth sub-eigenvalue to obtain a first eigenvalue.

[0049] In steps 10211 to 10216, the feature extraction layer includes five convolutional layers, which use 1D (one-dimensional) convolution to extract features in four dimensions: input voltage value, current value, operating frequency, and electrode spacing. At the same time, cross-layer connections are used for feature fusion to learn higher-dimensional features.

[0050] Among them, the first convolution layer is a 1D convolution, which outputs 32 feature maps with an output shape of [1, 32]; The second convolutional layer is a 1D convolution, which outputs 64 feature maps with an output shape of [1, 64]; The third convolutional layer is a 1D convolution, which outputs 128 feature maps with an output shape of [1, 128]; The fourth convolutional layer is a 1D convolution, which outputs 64 feature maps with an output shape of [1, 64]; The fifth convolutional layer is a 1D convolution, which outputs 32 feature maps with an output shape of [1, 32].

[0051] First, the first eigenvalue is input into the first convolutional layer for feature extraction to obtain a first sub-eigenvalue; the first eigenvalue is input into the second convolutional layer for feature extraction to obtain a second sub-eigenvalue; the second eigenvalue is input into the third convolutional layer for feature extraction to obtain a third sub-eigenvalue.

[0052] Then, the first convolution layer and the third convolution layer are connected, that is, the first sub-eigenvalue (32 features) and the third sub-eigenvalue (128 features) are concatenated by concat to obtain the fourth sub-eigenvalue, and the output shape is [1,160]. The fourth sub-eigenvalue is input into the fourth convolution layer for feature extraction to obtain the fifth sub-eigenvalue.

[0053] Then, the second convolutional layer and the fourth convolutional layer are connected, that is, the second sub-eigenvalue (64 features) and the fifth sub-eigenvalue (64 features) are concatenated by concat to obtain the first eigenvalue, and the output shape is [1, 128].

[0054] As an optional embodiment, the loss function of the first model during training includes a first error term, a second error term and a regularization term; The first error term represents the mean square error between the predicted plasma stability value and the actual plasma stability value; the second error term represents the mean square error between the predicted system energy consumption value and the actual system energy consumption value; the regularization term is used to control the physical rationality of the circuit operating parameters.

[0055] In an embodiment of the present invention, the loss function of the first model is a weighted multi-objective loss function, which combines the loss terms of multiple objectives such as plasma stability, energy consumption optimization and physical rationality of circuit parameters, and achieves a balance among different objectives in a weighted manner.

[0056] The loss function is as follows:

[0057] in, is the first error term, representing the mean square error between the predicted plasma stability value and the true plasma stability value; represents the second error term, which represents the mean square error between the predicted system energy consumption value and the actual system energy consumption value; Represents a regularization term, which is used to control the physical rationality of the circuit operating parameters (the input voltage value of the boost circuit, the current value in the circuit, the operating frequency of the circuit, and the electrode spacing). The regularization term can avoid over-optimization of a certain parameter.

[0058] During the model training process, based on the designed weighted multi-objective loss function, the loss function is minimized by the gradient descent method. The model will continuously update the weights of the network during the training process. After each forward propagation calculation of the prediction result, the loss function evaluates the difference between the predicted value and the actual value, calculates the gradient through the back propagation algorithm, and updates the network parameters. As the training progresses, the network will more and more accurately predict the impact of circuit parameters on stability and energy consumption.

[0059] The optimization algorithm used in the first model training process is the Adam optimizer. The Adam optimizer shows good performance during the training process and can automatically adjust the learning rate, thereby accelerating convergence and improving training stability.

[0060] After training the initial first model, the first model is obtained. During the training process, the input voltage (V), current (I), frequency (f), and electrode spacing (d) are first standardized and normalized and then passed to the input layer of the initial first model. The first to fifth convolutional layers extract low-level to high-level features from the input circuit parameters. Through the 1D convolution operation, the network will identify the relationship and pattern between different parameters. Cross-layer connection is to splice the output of the previous convolutional layer and integrate information at different levels, which helps to enhance the expressiveness of features, retain more information, and reduce information loss.

[0061] The self-attention mechanism calculates the relationship between input features and dynamically adjusts the importance of each feature to help the model capture the global dependencies between different circuit parameters. The weighted features output by the self-attention mechanism layer will enter the fully connected layer. Through the fully connected layer, the network converts the features from high dimensions to low dimensions. The features processed by the fully connected layer will be passed to the output layer. The output layer contains 2 neurons, corresponding to the stability of the plasma (S) and the energy consumption of the system (E).

[0062] As an optional embodiment, step 103 includes: Step 1031, initialize the particle swarm parameters; the particle swarm parameters include at least the number of particles, the position range and speed range of each particle, the inertia weight of the particle, the initial stability weight and the initial energy consumption weight; the position range is from the minimum value to the maximum value of the circuit operation parameter.

[0063] Step 1032: using the initial circuit operation parameters as initial position parameters of each particle; Step 1033: randomly generate an initial position and an initial speed of each particle in an initial state within the position range and the speed range, and generate an initial individual optimal position and an initial global optimal position of each particle; Step 1034, calculating the initial fitness of each particle according to the initial stability weight, the initial energy consumption weight, the initial position parameter, the initial system energy consumption value, the initial stability value and a preset fitness function; Step 1035: updating the speed of the particles within the speed range according to the particle swarm parameters and a preset speed update function to obtain an updated speed; Step 1036: updating the position of the particle within the position range according to the update speed and the initial position of the particle to obtain an updated position; Step 1037: Calculate the new fitness of the particle based on the position parameter of the particle at the updated position; Step 1038: If the new fitness is greater than the individual historical optimal value, the new fitness value is used as the current individual optimal value; if the new fitness is greater than the global historical optimal value, the new fitness value is used as the current global optimal value; Step 1039: When the preset maximum number of iterations is reached or the fitness reaches a preset convergence threshold, stop the calculation, obtain the position parameters of the particle located at the global optimal value, and obtain the optimal circuit parameters; the optimal circuit parameters are the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value.

[0064] In steps 1031 to 1039, first, the following parameter ranges and targets are defined: Voltage V: input voltage of the circuit, ranging from 100V to 1000V; Current I: operating current of the circuit, ranging from 1A to 10A; Frequency f: operating frequency of the circuit, ranging from 50Hz to 200Hz; Electrode spacing d: distance between electrodes, ranging from 1mm to 10mm.

[0065] The goal of algorithm optimization is to minimize energy consumption and maximize the stability of the plasma beam by optimizing these circuit parameters. Therefore, it is necessary to design a fitness function to evaluate the performance of each combination of circuit parameters.

[0066] The fitness function is the core of PSO optimization, which determines the performance of each particle's state parameters (i.e., circuit parameter combination). In this scheme, the fitness function considers two main factors: Stability of plasma beam: Generally, there is a strong correlation between stability and voltage and current, which can be quantified through simulation calculations or experimental data; Energy consumption: The energy consumption of a circuit is usually proportional to the product of voltage and current.

[0067] Then, the particle swarm parameters are initialized.

[0068] Specifically, the following particle swarm parameters can be set: Number of particles N: The number of particles affects the comprehensiveness of the search and the convergence speed. Set it to 50.

[0069] Dimension D: The dimension of each particle corresponds to the optimization variable, which is 4 (voltage V, current I, frequency f, electrode spacing d).

[0070] Particle position: The particle position ranges from the minimum and maximum values ​​of the circuit parameters.

[0071] Voltage V range: 100V to 1000V.

[0072] Current I range: 1A to 10A.

[0073] Frequency f range: 50Hz to 200Hz.

[0074] Electrode spacing d range: 1mm to 10mm.

[0075] Speed ​​Range: Controls the variation of particles in the search space, set to 10%-20%.

[0076] Inertia weight: The initial value is set to 0.9 and gradually decays to 0.4. The strategy of inertia weight decay adopts adaptive adjustment and dynamically adjusts according to the fitness improvement speed.

[0077] Individual learning factor and group learning factor: Set to 2.0, indicating that the weights of individuals and groups are balanced. In the early stages of the search, give the individual learning factor a higher weight to promote extensive exploration; in the convergence stage, gradually increase the weight of the group learning factor to help the algorithm find the optimal solution.

[0078] Weight coefficients of stability and energy consumption: In the initial stage, the weight coefficients are set to α =1.0, β =0.5, focusing on stability optimization; as the algorithm iterates, it is gradually adjusted to α =0.5, β =1.0, more attention is paid to energy consumption optimization.

[0079] Maximum number of iterations: 300.

[0080] Fitness convergence threshold: 0.0005.

[0081] The state of each particle represents a set of circuit parameters (i.e., voltage, current, frequency, and electrode spacing), and the initial circuit operation parameters are used as the initial position parameters of each particle. Within the position range and speed range, the initial position and initial speed of each particle in the initial state are randomly generated, and the initial individual optimal position and initial global optimal position of each particle are generated. The performance of these parameter combinations is then evaluated through the fitness function to obtain the initial fitness.

[0082] According to the particle swarm parameters and the preset speed update function, the particle speed is updated within the speed range to obtain the updated speed; according to the update speed and the initial position of the particle, the particle position is updated within the position range to obtain the updated position.

[0083] Based on the position parameters of the particle at the updated position (i.e., voltage, current, frequency, and electrode spacing), the new fitness of the particle is calculated. If the new fitness is greater than the individual historical optimal value , then update , that is, the new fitness value is taken as the current individual optimal value; if the new fitness value is greater than the global historical optimal value , then update , that is, the new fitness value is taken as the current global optimal value.

[0084] When the preset maximum number of iterations is reached or the fitness reaches the preset convergence threshold, the calculation is stopped, the position parameters of the particles at the global optimal value are obtained, and the optimal circuit parameters, that is, the optimal input voltage value, the optimal current value, the optimal operating frequency and the optimal electrode spacing are obtained. The optimal circuit parameters are the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value.

[0085] As an optional embodiment, step 1035 includes: Step 10351, calculating a first difference between the initial individual optimal position and the initial position of the particle, and a second difference between the initial global optimal position and the initial position; Step 10352: Determine the product of the first difference, the individual learning factor, and a preset first random number to obtain a first product; and determine the product of the second difference, the global learning factor, and a preset second random number to obtain a second product; Step 10353: determine the product of the inertia weight and the initial velocity of the particle to obtain a third product; Step 10354: determine the sum of the first product, the second product, and the third product to obtain an initial update speed; Step 10355: If the initial update speed is within the speed range, the initial update speed is used as the update speed.

[0086] In steps 10351-10355, the speed update function is as follows:

[0087] in, represents the speed at time t+1, represents the speed at time t. W represents the inertia weight, represents the optimal position of the initial individual, represents the initial position of the particle, represents the initial global optimal position, represent the individual learning factor and the group learning factor respectively, 、 Represent the first random number and the second random number respectively, and the range of the first random number and the second random number is [0, 1].

[0088] Time t can represent the initial time, then Indicates the initial speed Indicates the initial update speed. Determine whether the initial update speed is within the preset speed range. If so, use the initial update speed as the update speed. If not, use the maximum speed within the speed range as the update speed.

[0089] In addition, an adaptive update strategy is used to adjust the particle speed update amplitude. Specifically, the particle update amplitude is adjusted according to the stability change of the circuit parameters. If the stability of the circuit parameters improves, the particle speed is increased to conduct a wider search; if the stability of the circuit parameters decreases, the particle speed is reduced to avoid over-adjustment.

[0090] As an optional embodiment, step 1036 includes: According to the update speed and the initial position of the particle, and a preset position update function, the position of the particle is updated within the position range to obtain an updated position.

[0091] The position update function is as follows:

[0092] in, Indicates the update location. represents the initial position, Indicates the update speed.

[0093] The position is limited to the preset position range to ensure that the particle's updated position is limited to the allowable range of voltage, current, frequency and electrode spacing. The position update will be corrected in each iteration to ensure that the particle does not go out of the search space.

[0094] As an optional embodiment, step 1034 includes: Step 10341, inputting the initial position parameters into the first model to obtain an initial plasma stability value and an initial system energy consumption value; Step 10342: Substitute the initial position parameter into a preset regularization term to obtain a regularization value; Step 10343: determine the product of the initial stability weight and the initial plasma stability value to obtain a stability product; and determine the product of the initial energy consumption weight and the initial system energy consumption value to obtain an energy consumption product; Step 10344: determine the difference between the energy consumption product and the stability product, and determine the sum of the difference and the regularization value, and use the sum as the initial fitness of the particle.

[0095] In step 10341-step 10344, the initial position parameters are first input into the first model, and the first model outputs an initial plasma stability value and an initial system energy consumption value.

[0096] Then, the initial fitness of each particle is calculated using the initial stability weight, initial energy consumption weight, initial position parameter, initial system energy consumption value, initial stability value and preset fitness function.

[0097] Specifically, the fitness function is expressed as a combination of the system energy consumption E and the stability S of the plasma beam. The goal is to maximize S and minimize E. Therefore, the fitness function can be expressed as:

[0098] in, is a regularization term used to control circuit parameters (voltage , Current ,frequency , electrode spacing This regularization term can avoid over-optimization of a certain parameter.

[0099] E represents the system energy consumption, and S represents the stability of the plasma.

[0100] and is the weight coefficient of stability and energy consumption, and γ is the weight coefficient of the regularization term, which controls the influence of the constrained parameters on the fitness function.

[0101] As an optional embodiment, after step 1038, the method further includes: The stability weight and the energy consumption weight are dynamically adjusted according to the number of iterations.

[0102] In the embodiment of the present invention, the weights of stability and energy consumption are dynamically adjusted according to the number of iterations. In the early stage of iteration, more attention is paid to stability, that is, the weight of stability is greater than the weight of energy consumption; in the later stage of iteration, energy consumption optimization is gradually turned to, that is, the weight of energy consumption is greater than the weight of stability.

[0103] This solution uses an adaptive multi-objective particle swarm optimization algorithm to dynamically optimize circuit parameters while considering stability, energy consumption, and physical constraints. The real-time feedback mechanism and adaptive particle update strategy in the algorithm can be adjusted according to the real-time behavior of the circuit, thereby improving the optimization efficiency and the quality of the solution. By dynamically adjusting the weight of the fitness function, the optimization process can give priority to stability or energy consumption at different stages, and finally obtain the optimal circuit parameters that meet the requirements.

[0104] In summary, the circuit operation parameter adjustment method provided by this scheme realizes real-time optimization of circuit parameters through the combination of deep neural network simulation and particle swarm optimization algorithm. This method can significantly improve the working efficiency of equipment using plasma heating, reduce energy consumption, and improve the stability and output quality of plasma beams, and has great application value in the field of equipment using plasma heating.

[0105] Figure 2 1 is a structural block diagram of a circuit operation parameter adjustment device provided by an embodiment of the present invention. Figure 2 As shown, the device 200 includes: The parameter acquisition module 201 is used to acquire the initial circuit operation parameters of the device using plasma heating when it is in working state, and the initial circuit operation parameters include: the input voltage value of the boost circuit, the current value in the circuit, the operating frequency of the circuit and the electrode spacing; the electrode spacing represents the spacing between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating; the device using plasma heating refers to a device using the plasma heating principle as a heat source; A calculation module 202 is used to input the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, so as to obtain an initial stability value and an initial system energy consumption value of the circuit; The target parameter determination module 203 is used to use the initial circuit operation parameters as the initial position parameters of each particle, and perform fitness calculation according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm to obtain the target circuit operation parameters that maximize the plasma stability value and minimize the system energy consumption value; The adjustment module 204 is configured to adjust the initial circuit operating parameters to the target circuit operating parameters.

[0106] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0107] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0108] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that comply with the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0109] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for adjusting circuit operation parameters, characterized in that: The method comprises: Acquiring initial circuit operating parameters of the device using plasma heating in a working state, the initial circuit operating parameters including: an input voltage value of a boost circuit, a current value in a circuit, an operating frequency of the circuit, and an electrode spacing; the electrode spacing represents the spacing between a discharge electrode and a ground electrode of a plasma generator of the device using plasma heating; the device using plasma heating is a device that converts electrical energy into a high-temperature plasma flame using a plasma heating principle; Inputting the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, and obtaining an initial stability value and an initial system energy consumption value of the circuit; The initial circuit operation parameters are used as initial position parameters of each particle, and fitness calculation is performed according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm to obtain target circuit operation parameters that maximize the plasma stability value and minimize the system energy consumption value; The initial circuit operating parameters are adjusted to the target circuit operating parameters.

2. The method according to claim 1, characterized in that The first model includes a feature extraction layer, a feature fusion layer, a self-attention mechanism layer and a fully connected layer. The initial circuit operation parameters are input into the preset first model to calculate the plasma stability and system energy consumption, and the initial stability value and initial system energy consumption value of the circuit are obtained, including: Inputting the initial circuit operation parameters into the feature extraction layer and the feature fusion layer to perform initial feature extraction and feature fusion to obtain a first feature value; Input the first eigenvalue into the self-attention mechanism layer, adaptively calculate the target feature weight corresponding to each feature in the first eigenvalue by combining maximum pooling and average pooling, and weight the first eigenvalue by the target feature weight to obtain a second eigenvalue; Inputting the second eigenvalue into the feature extraction layer for further feature extraction to obtain a third eigenvalue; The third eigenvalue is input into a fully connected layer for full connection processing, and an initial stability value and an initial system energy consumption value of the circuit are output.

3. The method according to claim 2, characterized in that The feature extraction layer includes a first convolution layer, a second convolution layer, a third convolution layer, a fourth convolution layer and a fifth convolution layer, and the initial circuit operation parameters are input into the feature extraction layer and the feature fusion layer to perform initial feature extraction and fusion to obtain a first feature value, including: Inputting the first eigenvalue into the first convolutional layer for feature extraction to obtain a first sub-eigenvalue; Inputting the first eigenvalue into the second convolutional layer for feature extraction to obtain a second sub-eigenvalue; Inputting the second eigenvalue into the third convolutional layer for feature extraction to obtain a third sub-eigenvalue; Concatenate the first sub-eigenvalue and the third sub-eigenvalue to obtain a fourth sub-eigenvalue; Inputting the fourth sub-eigenvalue into the fourth convolutional layer for feature extraction to obtain a fifth sub-eigenvalue; The second sub-eigenvalue and the fifth sub-eigenvalue are concatenated to obtain a first eigenvalue.

4. The method according to claim 1, characterized in that: The loss function of the first model during training includes a first error term, a second error term and a regularization term; The first error term represents a mean square error between a predicted plasma stability value and a true plasma stability value; The second error term represents the mean square error between the predicted system energy consumption value and the actual system energy consumption value; The regularization term is used to control the physical rationality of the circuit operation parameters.

5. The method according to claim 1, characterized in that The initial circuit operation parameters are used as initial position parameters of each particle, and fitness calculation is performed according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm, including: Initialize particle swarm parameters; the particle swarm parameters include at least the number of particles, the position range and speed range of each particle, the inertia weight of the particle, the initial stability weight and the initial energy consumption weight; the position range is from the minimum value to the maximum value of the circuit operation parameter; Using the initial circuit operation parameters as initial position parameters of each particle; Randomly generate an initial position and an initial speed of each particle in an initial state within the position range and the speed range, and generate an initial individual optimal position and an initial global optimal position of each particle; Calculating the initial fitness of each particle according to the initial stability weight, the initial energy consumption weight, the initial position parameter, the initial system energy consumption value, the initial stability value and a preset fitness function; According to the particle group parameters and a preset speed update function, the speed of the particles is updated within the speed range to obtain an updated speed; According to the update speed and the initial position of the particle, updating the position of the particle within the position range to obtain an updated position; Calculating a new fitness of the particle based on the position parameter of the particle at the updated position; If the new fitness is greater than the individual historical optimal value, the new fitness value is used as the current individual optimal value; if the new fitness is greater than the global historical optimal value, the new fitness value is used as the current global optimal value; When the preset maximum number of iterations is reached or the fitness reaches a preset convergence threshold, the calculation is stopped, the position parameters of the particles located at the global optimal value are obtained, and the optimal circuit parameters are obtained; the optimal circuit parameters are the target circuit operating parameters that maximize the plasma stability value and minimize the system energy consumption value.

6. The method according to claim 5, characterized in that The updating of the particle speed within the speed range according to the initial parameters and a preset speed update function to obtain an updated speed includes: Calculating a first difference between the initial individual optimal position and the initial position of the particle, and a second difference between the initial global optimal position and the initial position; Determine the product of the first difference, the individual learning factor, and a preset first random number to obtain a first product; and determine the product of the second difference, the global learning factor, and a preset second random number to obtain a second product; determining the product of the inertia weight and the initial velocity of the particle to obtain a third product; Determine the sum of the first product, the second product, and the third product to obtain an initial update speed; If the initial update speed is within the speed range, the initial update speed is used as the update speed.

7. The method according to claim 5, characterized in that The step of calculating the initial fitness of each particle according to the initial stability weight, the initial energy consumption weight, the initial position parameter, the initial system energy consumption value, the initial stability value and a preset fitness function comprises: Inputting the initial position parameters into the first model to obtain an initial plasma stability value and an initial system energy consumption value; Substituting the initial position parameter into a preset regularization term to obtain a regularization value; Determine the product of the initial stability weight and the initial plasma stability value to obtain a stability product; and determine the product of the initial energy consumption weight and the initial system energy consumption value to obtain an energy consumption product; The difference between the energy consumption product and the stability product is determined, and the sum of the difference and the regularization value is determined, and the sum is used as the initial fitness of the particle.

8. A circuit operation parameter adjustment device, characterized in that: The device comprises: A parameter acquisition module, used to acquire initial circuit operation parameters of the device using plasma heating when it is in working state, wherein the initial circuit operation parameters include: input voltage value of the boost circuit, current value in the circuit, operating frequency of the circuit and electrode spacing; the electrode spacing represents the spacing between the discharge electrode and the ground electrode of the plasma generator of the device using plasma heating; the device using plasma heating refers to a device using the plasma heating principle as a heat source; A calculation module, used for inputting the initial circuit operation parameters into a preset first model to calculate plasma stability and system energy consumption, so as to obtain an initial stability value and an initial system energy consumption value of the circuit; a target parameter determination module, configured to use the initial circuit operation parameters as the initial position parameters of each particle, perform fitness calculation according to the initial position parameters, the initial stability value, the initial system energy consumption value, the first model and a preset adaptive multi-objective particle swarm optimization algorithm, and obtain the target circuit operation parameters that maximize the plasma stability value and minimize the system energy consumption value; The adjustment module is used to adjust the initial circuit operation parameters to the target circuit operation parameters.

9. A device using plasma heating, characterized in that: include: Controller; a memory for storing instructions executable by the controller; Wherein, the controller is configured to execute the instructions to implement the circuit operation parameter adjustment method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a controller of a device using plasma heating, the device using plasma heating is enabled to perform the circuit operation parameter adjustment method according to any one of claims 1 to 7.

Images