CN116961404A

CN116961404A - Control method, device, computer equipment and storage medium of resonant converter

Info

Publication number: CN116961404A
Application number: CN202310924736.7A
Authority: CN
Inventors: 李梦璇
Original assignee: Suzhou Inspur Intelligent Technology Co Ltd
Current assignee: Suzhou Inspur Intelligent Technology Co Ltd
Priority date: 2023-07-26
Filing date: 2023-07-26
Publication date: 2023-10-27

Abstract

The invention relates to the technical field of intelligent control, and discloses a control method and device of a resonant converter, computer equipment and a storage medium, wherein the method comprises the following steps: obtaining a reference input value and an actual output value corresponding to the current resonant converter; obtaining a correlation coefficient of the controller according to the reference input, the actual output value and the neural network, wherein the correlation coefficient comprises one or more of a proportional coefficient, an integral coefficient and a differential coefficient; generating a control quantity according to the correlation coefficient, the reference input, the actual output value and the controller; the resonant converter is controlled in accordance with the control quantity. The invention solves the problem that the proportional coefficient, the integral coefficient and the differential coefficient cannot be adaptively adjusted when the PID controller is used for controlling the resonant converter.

Description

Control method and device of resonant converter, computer equipment and storage medium

Technical Field

The present invention relates to the field of intelligent control technologies, and in particular, to a method and apparatus for controlling a resonant converter, a computer device, and a storage medium.

Background

At present, the traditional LLC (resonant conversion circuit) resonant converter is widely used in various application occasions by virtue of the advantages of high conversion efficiency, low electromagnetic interference, easiness in realizing magnetic integration and the like, but a plurality of problems still exist in the application process of the LLC resonant converter, such as lack of an intelligent control algorithm of the LLC resonant converter and the like. LLC resonant converter is a multivariable input, strong inter-variable coupling, nonlinear and time-varying system, and good control effect is difficult to realize by using the traditional classical control theory. Currently, conventional closed-loop control algorithms are mostly used for controlling LLC resonant converters, such as PID (proportional-integral-derivative) controllers. In order to achieve a better control effect, the PID controller must adjust the coefficients of the proportion, the integral and the derivative, however, when the PID controller is used to control the resonant converter, the proportional coefficient, the integral coefficient and the derivative coefficient cannot be adaptively adjusted, and thus a good control effect is difficult to achieve.

Therefore, the prior art has the problem that the adaptive adjustment of the proportional coefficient, the integral coefficient and the differential coefficient cannot be performed when the PID controller is used for controlling the resonant converter.

Disclosure of Invention

In view of the above, the present invention provides a control method, apparatus, computer device and storage medium for a resonant converter, so as to solve the problem that the proportional coefficient, integral coefficient and differential coefficient cannot be adaptively adjusted when the PID controller is used to control the resonant converter.

In a first aspect, the present invention provides a method for controlling a resonant converter, the method comprising:

obtaining a reference input value and an actual output value corresponding to the current resonant converter;

obtaining a correlation coefficient of the controller according to the reference input, the actual output value and the neural network, wherein the correlation coefficient comprises one or more of a proportional coefficient, an integral coefficient and a differential coefficient;

generating a control quantity according to the correlation coefficient, the reference input, the actual output value and the controller;

the resonant converter is controlled in accordance with the control quantity.

According to the control method of the resonant converter, the neural network is utilized to obtain the correlation coefficient of the controller, and the controller is adaptively adjusted, so that the controller can adapt to the working condition change of the resonant converter. The controller obtains control quantity according to the correlation coefficient, the reference input and the actual output value, and controls the resonant converter according to the control quantity, so as to realize intelligent self-adaptive control of the resonant converter. The problem that the proportional coefficient, the integral coefficient and the differential coefficient cannot be adaptively adjusted when the PID controller is used for controlling the resonant converter in the related technology is solved.

In an alternative embodiment, before obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network, the method further includes:

obtaining a training sample;

training the initial neural network according to the training sample and the neural network optimization algorithm to obtain target weights and target thresholds, wherein the target weights are model parameters of the initial neural network, and the target thresholds are used for judging whether to execute the next neuron in the initial neural network;

and adjusting the initial neural network according to the target weight and the target threshold value to obtain the neural network.

In the embodiment, the training sample and the neural network optimization algorithm are utilized to train the initial neural network, the trained neural network is obtained, the correlation coefficient of the controller can be adjusted according to the self-learning of the neural network, the controller can adapt to the working condition change of the resonant converter, and intelligent self-adaptive control is carried out on the resonant converter.

In an alternative embodiment, training the initial neural network according to the training sample and the neural network optimization algorithm to obtain the target weight and the target threshold value includes:

according to the training sample, obtaining input parameters and target output results of the initial neural network;

inputting the input parameters into an initial neural network and carrying out forward propagation to obtain an output result;

calculating an error between the output result and the target output result;

and carrying out back propagation on the error, and adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the error and a neural network optimization algorithm to obtain a target weight and a target threshold.

In this embodiment, input parameters in a training sample are input into an initial neural network to obtain an output result; calculating an error between the output result and the target output result; and adjusting the weight and the threshold of the initial neural network according to the error and the neural network optimization algorithm to obtain a target weight and a target threshold, and completing training of the neural network, so that the neural network output result is more accurate.

In an alternative embodiment, adjusting weights and thresholds of an output layer and a hidden layer in an initial neural network according to an error and a neural network optimization algorithm to obtain a target weight and a target threshold includes:

determining a learning rate and a training frequency threshold of a neural network optimization algorithm;

and adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the learning rate until the error is smaller than a preset threshold or reaches a training frequency threshold, stopping adjusting to obtain a target weight and a target threshold.

In this embodiment, the neural network optimization algorithm adjusts weights and thresholds of an output layer and a hidden layer in the initial neural network according to the learning rate to obtain a target weight and a target threshold, so that the neural network performance is optimal, and the output result is more accurate.

In an alternative embodiment, the obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network includes:

calculating a deviation between the reference input and the actual output value based on the reference input and the actual output value;

the reference input, the actual output value and the deviation are input into a neural network to obtain a correlation coefficient, wherein the correlation coefficient comprises a proportional coefficient, an integral coefficient and a differential coefficient.

In the present embodiment, the reference input, the actual output value, and the deviation are input to the neural network, and the neural network outputs the proportional coefficient, the integral coefficient, and the differential coefficient. The correlation coefficient of the controller is regulated by self-learning of the neural network so as to achieve the optimization of the control effect of the resonant converter.

and inputting the reference input value and the actual output value into a neural network to obtain a correlation coefficient.

In the present embodiment, the reference input and the actual output value are taken as inputs to the neural network, and the neural network outputs the correlation coefficient. The input and output conditions of the neural network are changed, and the application range of the invention is enriched.

In an alternative embodiment, generating the control quantity based on the correlation coefficient, the reference input, the actual output value, and the controller includes:

and obtaining the control quantity according to the correlation coefficient, the deviation and the actual output value calculation formula of the controller.

In the present embodiment, the control amount is calculated using the correlation coefficient, the deviation between the reference input and the actual output value, and the actual output value calculation formula. The control quantity is used for controlling the resonant converter, so that the steady-state performance and the dynamic effect of the resonant converter can be improved.

In a second aspect, the present invention provides a control device for a resonant converter, the device comprising:

the first acquisition module is used for acquiring reference input and actual output values corresponding to the current resonant converter;

the obtaining module is used for obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network, wherein the correlation coefficient comprises one or more of a proportional coefficient, an integral coefficient and a differential coefficient;

the generation module is used for generating a control quantity according to the correlation coefficient, the reference input, the actual output value and the controller;

and the control module is used for controlling the resonant converter according to the control quantity.

In a third aspect, the present invention provides a computer device comprising: the processor executes the computer instructions, thereby executing the control method of the resonant converter according to the first aspect or any embodiment corresponding to the first aspect.

In a fourth aspect, the present invention provides a computer-readable storage medium having stored thereon computer instructions for causing a computer to execute the control method of the resonant converter of the first aspect or any of the embodiments corresponding thereto.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a block diagram of an existing PID controller according to an embodiment of the invention;

FIG. 2 is a flow chart of a method of controlling a resonant converter according to an embodiment of the invention;

FIG. 3 is a block diagram of a BP neural network-based PID control system according to an embodiment of the invention;

fig. 4 is a schematic structural diagram of a BP neural network according to an embodiment of the present invention;

fig. 5 is a block diagram of a control device of a resonant converter according to an embodiment of the present invention;

fig. 6 is a schematic diagram of a hardware structure of a computer device according to an embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The requirements of various application scenes on power supplies are higher and higher, including: the power supply switching frequency is improved, the dynamic requirements on the power supply are more severe, the stability requirements on the power supply under a complex environment are improved, and the power supply product is required to have intelligent functions of fault diagnosis reporting, multi-module communication interconnection, power supply state monitoring and the like under more working conditions, so that the power supply technology has improved over the sky and over the ground in the past, and the digital power supply technology has developed to a great extent. Currently, LLC (resonant conversion circuit) resonant converters are widely used in industries such as new energy and server power supply by virtue of excellent characteristics that soft switching (switching loss of power devices is reduced, conversion efficiency is high), EMI (Electromagnetic Interference ) is small, and power class is large in a full load range. Although conventional LLC resonant converters are widely used in various applications, there are still a number of problems in the application process, such as lack of intelligent control algorithms for LLC resonant converters, etc. Therefore, the intelligent closed-loop control technology of the LLC resonant converter has great significance in deep research and improvement.

LLC resonant converter is a multivariable input, strong inter-variable coupling, nonlinear and time-varying system, and a good control effect is difficult to realize by a traditional closed-loop control algorithm. The closed-loop controller designed by using intelligent control algorithms such as a neural network and the like can improve the steady-state and dynamic response of the LLC resonant converter, and achieve intelligent self-adaptive control of the LLC resonant converter. Currently, conventional closed-loop control algorithms are mostly used for controlling LLC resonant converters, such as PID (proportional-integral-derivative) controllers.

PID (proportional, integral, derivative, proportional, integral, derivative) controllers are based on PID algorithms, a common "hold steady" control algorithm. The PID controller must adjust the values of the proportional, integral and derivative values to achieve a good control effect. As shown in fig. 1, r (t) is a reference input, y (t) is an actual output value, e (t) is a difference between the reference input and the actual output value, t is time, the PID controller includes three parts, i.e., proportional, integral and derivative, which are calculated according to e (t), a proportional coefficient of the controller, an integral coefficient of the controller and a derivative coefficient of the controller, respectively, and are summarized to obtain a control quantity u (t), and the control quantity u (t) is used to control a controlled object, which is an LLC resonant converter. The conventional PID controller cannot adaptively adjust the scaling factor, the integration factor and the differentiation factor. By utilizing any nonlinear expression capability of the neural network, optimal combination of proportion, integral and derivative can be realized through learning of system performance.

The embodiment of the invention provides a control method of a resonant converter, which enables a PID controller to be self-adaptive to the working condition change of an LLC resonant converter through self-learning adjustment of relevant parameters of the PID controller of a neural network so as to achieve intelligent self-adaptive control of the LLC resonant converter and improve the dynamic effect and the technical effect of steady-state performance of the LLC resonant converter.

According to an embodiment of the present invention, there is provided a control method embodiment of a resonant converter, it should be noted that the steps shown in the flowchart of the drawings may be performed in a smart terminal having data processing capability, and that although a logic sequence is shown in the flowchart, in some cases, the steps shown or described may be performed in a different sequence than here.

In this embodiment, a control method of a resonant converter is provided, which may be used in the above-mentioned intelligent terminal, and fig. 2 is a flowchart of a control method of a resonant converter according to an embodiment of the present invention, as shown in fig. 2, where the flowchart includes the following steps:

step S201, obtaining a reference input and an actual output value corresponding to the current resonant converter.

Specifically, the control method of the resonant converter combines the PID controller and the neural network, and the control method of the resonant converter is a method of the PID controller on the LLC resonant converter based on the neural network, wherein the LLC resonant converter is a controlled object.

And acquiring a reference input r (t) and an actual output value y (t) corresponding to the current resonant converter.

Step S202, obtaining a correlation coefficient of the controller according to the reference input, the actual output value and the neural network, wherein the correlation coefficient comprises one or more of a proportional coefficient, an integral coefficient and a differential coefficient.

Specifically, an artificial Neural network (Artificial Neural Network, ANN), simply called Neural Networks (Neural Networks), is a network structure that mimics the processing of information by the human brain, by weighting a plurality of input variables, and using nonlinear activation to produce nonlinear outputs. The neural network technology can be utilized to perform deep processing and extraction on the input information, and the neural network technology has the characteristics of high precision, high accuracy, high robustness, high intelligence and the like. With the improvement of the performance of the PID controller, the neural network can intelligently control the complex system in real time on the basis of the PID controller. The neural network comprises: RNN (Recurrent Neural Network) recurrent neural network, LSTM (Long Short-Term Memory network) recurrent neural network, BP (back propagation) neural network, etc., wherein the BP neural network is one of the most widely used neural networks, and the learning process consists of two processes of forward propagation of signals and backward propagation of errors, and can be iterated by using an error back propagation algorithm, so that the errors are gradually reduced to be within an acceptable range.

And inputting the reference input r (t) and the actual output value y (t) as input parameters into the neural network which is trained by using the training samples. Or firstly calculating a difference e (t) between a reference input r (t) and an actual output value y (t), then selecting any two or three of r (t), y (t) and e (t) as input parameters, and inputting the input parameters into a neural network to obtain a correlation coefficient. The neural network needs to be trained in advance with a large number of training samples, which need to contain the above-mentioned input parameters and corresponding correlation coefficients. The neural network outputs the correlation coefficient of the controller, and the correlation coefficient comprises the proportionality coefficient K of the controller _p Integral coefficient T _i Differential coefficient T _d Which may be set according to requirements, for example: the correlation coefficient includes an integral coefficient and a differential coefficient, and the proportional coefficient is set according to the test result. Alternatively, the correlation coefficient contains three coefficients.

In step S203, a control amount is generated according to the correlation coefficient, the reference input, the actual output value, and the controller.

Specifically, a difference e (t) between the reference input r (t) and the actual output value y (t) is calculated. Determining the proportional coefficient K of the controller according to the correlation coefficient _p Integral coefficient T _i Differential coefficient T _d Based on the difference e (t) and the proportionality coefficient K by using PID controller _p Calculating the proportional value according to the difference e (T) and the integral coefficient T _i Calculating the integral value according to the difference e (T) and the differential coefficient T _d The integrated value is calculated, and the control amounts u (t) are obtained by summarizing.

Step S204, controlling the resonant converter according to the control quantity.

Specifically, the controlled object is controlled by using the control quantity u (t), and the controlled object is an LLC resonant converter.

In some alternative embodiments, before deriving the correlation coefficient of the controller from the reference input, the actual output value, and the neural network, the method further comprises:

obtaining a training sample;

Specifically, according to a large number of tests, a large number of training samples are selected from the test data, wherein the training samples need to be selected according to input parameters and output parameters of the neural network, for example: if the input parameters of the neural network are: the reference input r (t), the actual output value y (t) and the difference e (t), and the output parameters are as follows: scaling factor K _p Integral coefficient T _i Differential coefficient T _d Each training sample should also contain the data described above.

Inputting the training sample into an initial neural network, and training the initial neural network by using a neural network optimization algorithm to obtain target weight and a target threshold, for example: gradient descent, adam (Adaptive Moment Estimation ) algorithm, and the like. And adjusting the initial neural network according to the target weight and the target threshold value to obtain the trained neural network.

Compared with a conventional PID controller, the method has the advantages that the controlled object is directly subjected to closed-loop control, and three parameters are in an online adjustment mode. In the embodiment, the parameter K of the PID controller is regulated through self-learning of the neural network _p 、T _i 、T _i To achieve optimization of performance metrics.

In some alternative embodiments, training the initial neural network according to the training sample and the neural network optimization algorithm to obtain the target weight and the target threshold value includes:

calculating an error between the output result and the target output result;

Specifically, the embodiment adopts a BP neural network self-correction PID controller to design an intelligent control algorithm, and provides a method for an LLC resonant converter based on the self-correction BP neural network PID controller. The input parameters of the BP neural network are as follows: the reference input r (t), the actual output value y (t) and the difference e (t), and the output parameters are as follows: scaling factor K _p Integral coefficient T _i Differential coefficient T _d For example, as shown in FIG. 3, the reference input r, the actual output value y, and the difference e are input into the BP neural network, and the BP neural network outputs the scaling factor K _p Integral coefficient T _i Differential coefficient T _d PID controller based on difference e and proportional coefficient K _p Integral coefficient T _i Differential coefficient T _d And generating a control quantity u, and controlling the controlled object by using the u. The whole network structure of the BP neural network comprises an input layer, one to multiple hidden layers and an output layer, as shown in fig. 4, after signals are input, the signals sequentially pass through the input layer, the hidden layers and the output layer to be transmitted forward, and an output result is obtained. And obtaining an error according to the output result and the expected signal, and carrying out error back transmission. The input layer is the first layer in the neural network, receives input signals and passes them on to the next layer. The hidden layer is called a hidden layer, and the hidden layer may be a plurality of layers, except for the input layer and the output layer. The hidden layer may perform various operations on signals passed in by the input layer. The output layer is the last layer of the network and outputs the result value predicted by the model.

The BP neural network model training process mainly comprises two stages, wherein the first stage is forward propagation of signals, passes through a hidden layer from an input layer and finally reaches an output layer. The forward propagation is to make the information enter the network from the input layer, and the information is calculated by each layer in turn, wherein the direct calculation mode is that the value of each layer is multiplied by the corresponding weight and bias variable (activation function) to obtain the final output layer result. The second stage is the counter-propagation of the error from the output layer to the hidden layer and finally to the input layer, the network parameters are adjusted by calculating the error between the output layer and the desired value, and the weight and bias of the hidden layer to the output layer and the weight and bias of the input layer to the hidden layer are sequentially adjusted, thereby minimizing the error.

Based on the above. And obtaining input parameters and target output results of the initial neural network according to the training samples. And inputting the input parameters into the initial neural network, carrying out forward propagation, and sequentially passing through the input layer, the hidden layer and the output layer to obtain an output result. Calculating an error between the output result and the target output result, carrying out back propagation on the error, and adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the error and a neural network optimization algorithm to obtain a target weight and a target threshold.

Compared with a conventional PID controller which directly performs closed-loop control on a controlled object, three parameters are in an on-line adjustment mode, and the PID parameter K is adjusted through self-learning of the BP neural network in the embodiment _p 、T _i 、T _i To achieve optimization of performance metrics.

In some alternative embodiments, adjusting weights and thresholds of an output layer and a hidden layer in an initial neural network according to an error and neural network optimization algorithm to obtain a target weight and a target threshold includes:

Specifically, the BP neural network adopts a gradient descent method to adjust the weight and the threshold value of each layer of nodes, so that the error is reduced. The gradient descent method is easy to sink into a trap with a local minimum value, so that the gradient descent method cannot descend to a point with the minimum gradient, and the performance of the BP algorithm is seriously affected. Therefore, the invention adopts Adam algorithm to optimize BP neural network. Adam's algorithm is an algorithm that combines Momentum's algorithm with RMSProp (root mean square propagation ) algorithm. The Momentum algorithm is based on the moving exponential weighted average of the gradient, can solve the problem of large update amplitude swing, and can enable the convergence speed of the network to be faster. The RMSProp algorithm calculates a differential square weighted average for the gradient, which is favorable for eliminating the direction of large swing amplitude, and is used for correcting the swing amplitude so that the swing amplitude of each dimension is smaller; on the other hand, the network function converges faster. Adam combines the advantages of the two, and the learning rate can be adaptively adjusted by using an Adam algorithm to accelerate convergence.

The neural network optimization algorithm optimizes the neural network by using the Adam algorithm, and combines the advantages of the self-adaptive learning rate optimization algorithm and the momentum method. The expression of Adam algorithm is z _t ＝β ₁ ×m _{t 1} +(1β ₁ ) X dk, where m _{t 1} And m _t First moment estimation before iteration and after iteration respectively, m ₁ For an exponentially weighted average parameter, dk represents the gradient of the weight w or threshold b, β ₁ To control the gradient average parameters. Similarly, the second moment of Adam's algorithm is estimated as: v _t ＝β ₂ ×v _{t 1} +(1β ₂ )×dk ² Wherein v is _{t 1} And v _t Respectively estimating the second moment before and after iteration, beta ₂ For exponentially weighted average parameters dk ² Representing the square of the gradient of the weight w or the threshold b.

The first-order momentum mt and the second-order momentum vt are combined together, the weight and the threshold value are updated, and the expression is as follows:

wherein w is _t 、b _t Is the adjusted weight and threshold, w _{t 1} 、b _{t 1} Is the weight and threshold to be trained, α is the learning rate, and θ is a constant. In order to prevent the phenomenon of over-fitting in the model training process, the model is regularized by adopting a Dropout method.

Based on the above, the learning rate of Adam algorithm is determined, for example: the learning rate was set to 0.0001. Determining a training frequency threshold of the Adam algorithm, for example: the maximum number of exercises (i.e., the threshold number of exercises) is taken as 100. And setting automatic termination and saving the optimal model in the training process. By using the method, the weights and the thresholds of the output layer and the hidden layer in the initial neural network are adjusted by combining the learning rate, and the adjustment is stopped until the error is smaller than a preset threshold or reaches the training frequency threshold, so as to obtain the target weight and the target threshold.

In some alternative embodiments, obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network includes:

Specifically, the input parameters of the neural network in this embodiment are: the reference input r (t), the actual output value y (t) and the difference e (t), and the output parameters are as follows: scaling factor K _p Integral coefficient T _i Differential coefficient T _d 。

A difference e (t) between the reference input r (t) and the actual output value y (t) is calculated. Inputting the reference input r (t), the actual output value y (t) and the difference e (t) into a neural network, and outputting a proportionality coefficient K by the neural network _p Integral coefficient T _i Differential coefficient T _d 。

Specifically, the input parameters of the present embodiment are the reference input r (t) and the actual output value y (t). The reference input r (t) and the actual output value y (t) are input to a neural network, which outputs a correlation coefficient.

In addition, the input parameters and output parameters of the neural network may be changed according to the need, and the input parameters are not limited to the reference input and actual output values.

In some alternative embodiments, generating the control amount based on the correlation coefficient, the reference input, the actual output value, and the controller includes:

Specifically, a difference e (t) between the reference input r (t) and the actual output value y (t) is calculated.

Determining the coefficients of the controller based on the correlation coefficients, comprising: scaling factor K _p Integral coefficient T _i Differential coefficient T _d . In addition, the correlation coefficient of the neural network output contains which coefficients can be set according to the requirement and can containOne or more of a proportional coefficient, an integral coefficient, and a differential coefficient. For example: the correlation coefficient comprises an integral coefficient and a differential coefficient, and the proportional coefficient is set according to the test result, and then the integral coefficient T is determined according to the correlation coefficient _i And differential coefficient T _d Scaling factor K _p And determining according to the setting content.

The difference e (t) and the proportionality coefficient K _p Integral coefficient T _i Differential coefficient T _d The actual output value calculation formula of the input controller, for example, formula (1), calculates the control amount u (t).

Wherein K is _p Is a proportional coefficient of the controller; t (T) _i Integrating the coefficients for the controller; t (T) _d Is the differential coefficient of the controller.

It should be noted that, at present, the intelligent control algorithm such as the neural network has little research in the DC-DC converter, especially in the application to the LLC converter, and the intelligent control algorithm such as the neural network is very suitable for being used as the closed-loop control algorithm of the complex topology such as the LLC converter, therefore, the invention has higher application type innovation value.

The present embodiment also provides a control device for a resonant converter, which is used to implement the foregoing embodiments and preferred embodiments, and is not described in detail. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

The present embodiment provides a control device for a resonant converter, as shown in fig. 5, including:

a first obtaining module 501, configured to obtain a reference input value and an actual output value corresponding to a current resonant converter;

an obtaining module 502, configured to obtain a correlation coefficient of the controller according to the reference input, the actual output value, and the neural network, where the correlation coefficient includes one or more of a proportional coefficient, an integral coefficient, and a differential coefficient;

a generating module 503, configured to generate a control amount according to the correlation coefficient, the reference input, the actual output value, and the controller;

a control module 504 for controlling the resonant converter according to the control quantity.

In some alternative embodiments, the apparatus further comprises:

the second acquisition module is used for acquiring training samples;

the training module is used for training the initial neural network according to the training sample and the neural network optimization algorithm to obtain target weight and a target threshold value, wherein the target weight is a model parameter of the initial neural network, and the target threshold value is used for judging whether to execute the next neuron in the initial neural network;

and the adjusting module is used for adjusting the initial neural network according to the target weight and the target threshold value to obtain the neural network.

In some alternative embodiments, the training module includes:

the first obtaining unit is used for obtaining input parameters and target output results of the initial neural network according to the training samples;

the second obtaining unit is used for inputting the input parameters into the initial neural network and carrying out forward propagation to obtain an output result;

a first calculation unit for calculating an error between the output result and the target output result;

and the adjusting unit is used for carrying out counter propagation on the error, and adjusting the weights and the thresholds of the output layer and the hidden layer in the initial neural network according to the error and the neural network optimization algorithm to obtain a target weight and a target threshold.

In some alternative embodiments, the adjustment unit comprises:

the determining submodule is used for determining the learning rate and the training frequency threshold value of the neural network optimization algorithm;

and the adjustment sub-module is used for adjusting the weights and the thresholds of the output layer and the hidden layer in the initial neural network according to the learning rate until the error is smaller than a preset threshold or reaches the training frequency threshold, and stopping adjustment to obtain the target weight and the target threshold.

In some alternative embodiments, the deriving module 502 includes:

a second calculation unit for calculating a deviation between the reference input and the actual output value based on the reference input and the actual output value;

and the third obtaining unit is used for inputting the reference input, the actual output value and the deviation into the neural network to obtain a correlation coefficient, wherein the correlation coefficient comprises a proportional coefficient, an integral coefficient and a differential coefficient.

In some alternative embodiments, the deriving module 502 includes:

and a fourth obtaining unit, configured to input the reference input value and the actual output value into the neural network, and obtain the correlation coefficient.

In some alternative embodiments, the generating module 503 includes:

a third calculation unit for calculating a deviation between the reference input and the actual output value based on the reference input and the actual output value;

and a fifth obtaining unit, configured to obtain the control amount according to the correlation coefficient, the deviation, and the actual output value calculation formula of the controller.

Further functional descriptions of the above respective modules and units are the same as those of the above corresponding embodiments, and are not repeated here.

The control means of the resonant converter in this embodiment are presented in the form of functional units, here referred to as ASIC (Application Specific Integrated Circuit ) circuits, processors and memories executing one or more software or fixed programs, and/or other devices that can provide the above described functionality.

The embodiment of the invention also provides computer equipment, which is provided with the control device of the resonant converter shown in the figure 5.

Referring to fig. 6, fig. 6 is a schematic structural diagram of a computer device according to an alternative embodiment of the present invention, as shown in fig. 6, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components are communicatively coupled to each other using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions executing within the computer device, including instructions stored in or on memory to display graphical information of the GUI on an external input/output device, such as a display device coupled to the interface. In some alternative embodiments, multiple processors and/or multiple buses may be used, if desired, along with multiple memories and multiple memories. Also, multiple computer devices may be connected, each providing a portion of the necessary operations (e.g., as a server array, a set of blade servers, or a multiprocessor system). One processor 10 is illustrated in fig. 6.

The processor 10 may be a central processor, a network processor, or a combination thereof. The processor 10 may further include a hardware chip, among others. The hardware chip may be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general-purpose array logic, or any combination thereof.

Wherein the memory 20 stores instructions executable by the at least one processor 10 to cause the at least one processor 10 to perform a method for implementing the embodiments described above.

The memory 20 may include a storage program area that may store an operating system, at least one application program required for functions, and a storage data area; the storage data area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, memory 20 may optionally include memory located remotely from processor 10, which may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

Memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk, or solid state disk; the memory 20 may also comprise a combination of the above types of memories.

The computer device also includes a communication interface 30 for the computer device to communicate with other devices or communication networks.

The embodiments of the present invention also provide a computer readable storage medium, and the method according to the embodiments of the present invention described above may be implemented in hardware, firmware, or as a computer code which may be recorded on a storage medium, or as original stored in a remote storage medium or a non-transitory machine readable storage medium downloaded through a network and to be stored in a local storage medium, so that the method described herein may be stored on such software process on a storage medium using a general purpose computer, a special purpose processor, or programmable or special purpose hardware. The storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, a solid state disk or the like; further, the storage medium may also comprise a combination of memories of the kind described above. It will be appreciated that a computer, processor, microprocessor controller or programmable hardware includes a storage element that can store or receive software or computer code that, when accessed and executed by the computer, processor or hardware, implements the methods illustrated by the above embodiments.

Although embodiments of the present invention have been described in connection with the accompanying drawings, various modifications and variations may be made by those skilled in the art without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope of the invention as defined by the appended claims.

Claims

1. A method of controlling a resonant converter, the method comprising:

and controlling the resonant converter according to the control quantity.

2. The method of claim 1, wherein prior to said deriving the correlation coefficient for the controller from the reference input, the actual output value, and the neural network, the method further comprises:

obtaining a training sample;

training an initial neural network according to the training sample and a neural network optimization algorithm to obtain a target weight and a target threshold, wherein the target weight is a model parameter of the initial neural network, and the target threshold is used for judging whether to execute the next neuron in the initial neural network;

3. The method according to claim 2, wherein training the initial neural network according to the training samples and the neural network optimization algorithm to obtain the target weight and the target threshold comprises:

obtaining input parameters and target output results of the initial neural network according to the training samples;

inputting the input parameters into the initial neural network and carrying out forward propagation to obtain an output result;

calculating an error between the output result and the target output result;

and carrying out counter propagation on the error, and adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the error and the neural network optimization algorithm to obtain the target weight and the target threshold.

4. The method of claim 3, wherein said adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the error and the neural network optimization algorithm to obtain the target weight and the target threshold comprises:

determining a learning rate and a training frequency threshold of the neural network optimization algorithm;

and adjusting weights and thresholds of an output layer and a hidden layer in the initial neural network according to the learning rate until the error is smaller than a preset threshold or reaches the training frequency threshold, stopping adjusting to obtain the target weight and the target threshold.

5. The method according to claim 1 or 2, wherein the obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network comprises:

calculating a deviation between the reference input and the actual output value from the reference input and the actual output value;

and inputting the reference input, the actual output value and the deviation into the neural network to obtain the correlation coefficient, wherein the correlation coefficient comprises a proportional coefficient, an integral coefficient and a differential coefficient.

6. The method according to claim 1 or 2, wherein the obtaining the correlation coefficient of the controller according to the reference input, the actual output value and the neural network comprises:

and inputting the reference input and the actual output value into the neural network to obtain the correlation coefficient.

7. The method of claim 1, wherein the generating a control quantity based on the correlation coefficient, the reference input, the actual output value, and the controller comprises:

and obtaining the control quantity according to the correlation coefficient, the deviation and an actual output value calculation formula of the controller.

8. A control device for a resonant converter, the device comprising:

the obtaining module is used for obtaining a correlation coefficient of the controller according to the reference input, the actual output value and the neural network, wherein the correlation coefficient comprises one or more of a proportional coefficient, an integral coefficient and a differential coefficient;

9. A computer device, comprising:

a memory and a processor communicatively coupled to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the method of controlling a resonant converter of any of claims 1 to 7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon computer instructions for causing a computer to execute the control method of the resonant converter according to any one of claims 1 to 7.