An event-triggered online particle swarm optimization algorithm based data-driven control method

Abstract

The application discloses a data-driven control method based on an event-triggered online particle swarm algorithm and belongs to the field of power electronic converter control methods.The method comprises the following steps: constructing a data-driven model of a power electronic converter based on partial-form dynamic linearization; adopting an improved event-triggered online particle swarm algorithm to estimate data-driven model parameters in real time; and designing a model-free predictive controller based on the data-driven model and a finite control set-model predictive control method.The application introduces a partial-form dynamic linearization data-driven model, eliminates the dependence of a traditional finite control set-model predictive control method on prior information of a model, simultaneously introduces an event-triggering mechanism to control the activation of an online particle swarm algorithm, reduces the calculation complexity on the basis of sufficient control precision, and further improves the real-time performance of the control method.

Description

Technical Field

[0001] This invention relates to the field of power electronic converter control methods, specifically to a data-driven control method based on an event-triggered online particle swarm optimization algorithm. Background Technology

[0002] Finite control set-model predictive control inherits the multi-objective constraint optimization capability and excellent dynamic performance of model predictive control. This method fully considers the discrete characteristics and finite switching states of the power converter and has been widely used. However, model dependency problem still severely restricts the performance of traditional finite control set-model predictive control in practical applications.

[0003] In practical applications, parameter mismatch or drift is unavoidable, significantly affecting the performance of finite control set model predictive control in power electronic converter systems. Severe parameter mismatch can even lead to system instability. Therefore, a method is urgently needed to address the model dependency problem of traditional finite control set model predictive control. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention constructs a partial scheme dynamic linearized data-driven model for power electronic converters, uses an improved event-triggered online particle swarm optimization algorithm to estimate model parameters in real time, and performs converter control based on the dynamic data model through finite control set-model predictive control, thereby solving the model dependency problem of traditional finite control set-model predictive control.

[0005] To achieve the above objectives, this invention provides a data-driven control method based on an event-triggered online particle swarm optimization algorithm, comprising the following steps: (1) Construct a data-driven model for a power electronic converter based on partial scheme dynamic linearization; (2) An improved event-triggered online particle swarm optimization algorithm is used to estimate data-driven model parameters in real time; (3) Design a model-free predictive controller based on data-driven model and finite control set-model predictive control method.

[0006] Furthermore, the partial scheme dynamic linearized power electronic converter data-driven model is specifically as follows: in: Represents the depth of the data involved in the computation. >0; It is the change in the system state of the power electronic converter in the next sampling period; It controls the input increment vector; It is the change in control input within a sampling period; It is the parameter matrix of the data-driven model.

[0007] Furthermore, the improved event-triggered online particle swarm optimization algorithm specifically involves: using data-driven model parameters as particle positions in the online particle swarm optimization algorithm, and using the real-time system state of the power electronic converter as the basis for particle evaluation; comprehensively evaluating particles using the tracking errors of the power electronic converter system over multiple sampling periods, thereby guiding the estimation of data-driven model parameters; and introducing an event-triggered mechanism, which activates the online particle swarm optimization algorithm if the tracking error of the current globally optimal particle exceeds a preset threshold, otherwise maintaining the data model parameters unchanged.

[0008] Furthermore, the steps for updating model parameters using the event-triggered online particle swarm optimization algorithm are as follows: (2.1) Apply particles as data-driven model parameters; (2.2) Collect the power electronic converter system state for multiple sampling periods and calculate the cost of the current particle; (2.3) Repeat steps (2.1) and (2.2) until all particles have been evaluated; (2.4) Update the current global optimum and individual optimum, and update the population position; (2.5) Apply the globally optimal particle as the data-driven model parameter; (2.6) Collect the power electronic converter system state for multiple sampling periods and calculate the cost of the current global optimal particle; (2.7) Determine whether to continue using the online particle swarm optimization algorithm for model parameter estimation based on the event triggering conditions; if the event triggering condition is met, update the data model parameters; if the event triggering condition is not met, keep the data model parameters unchanged. (2.8) Repeat steps (2.1)-(2.7) to adjust the transformer model parameters in real time using the online particle swarm algorithm.

[0009] Furthermore, the cost function of the event-triggered online particle swarm optimization algorithm is designed as follows: in: This represents the expected output of the power electronic converter system in the current sampling period; Represents the current state of the system; This indicates the number of sampling periods used to evaluate the particles.

[0010] Furthermore, the event triggering mechanism of the event-triggered online particle swarm optimization algorithm is as follows: in: It is the event trigger threshold. >0; and These represent the data-driven model parameter estimates for the current and previous time points, respectively. It is particles in In the next evaluation cycle, the first The value of the next generation; It is the total value of an evaluation of a particle.

[0011] Furthermore, step (3) specifically involves: (3.1) Calculate the finite control set based on the power electronic converter system; (3.2) Predict the set of states of a finite control set based on the current data-driven model; (3.3) Calculate the cost of all control inputs based on the predicted state set; (3.4) Sort the cost values, determine the control input with the smallest cost value, and apply it as the control input to the power electronic converter.

[0012] Furthermore, the cost function of the model-free predictive controller is specifically as follows: in: This represents the expected output of the power electronic converter system in the next sampling period; It is the system state at the next moment obtained from the current model-free predictive controller.

[0013] The beneficial effects of this invention are: 1. This invention uses a partial scheme dynamic linearized data-driven model as the system model. This model only depends on the system's input and output data, eliminating the dependence of traditional finite control set-model predictive control methods on prior model information.

[0014] 2. This invention uses an event-triggered online particle swarm optimization algorithm to estimate data-driven model parameters in real time, thereby addressing parameter mismatch and drift and maintaining superior control performance under different operating conditions.

[0015] 3. This invention introduces an event-triggered mechanism to control the activation of the online particle swarm optimization algorithm, thereby reducing computational complexity while maintaining sufficient control accuracy and improving the real-time performance of the control method.

[0016] 4. This invention constructs a model-free predictive controller based on a data-driven model and a finite control set-model predictive control. The controller parameters (model parameters) are dynamically adjusted according to the real-time state of the power electronic converter, thereby ensuring the control effect of the converter. Attached Figure Description

[0017] Figure 1 For having A topology diagram of a three-phase two-level inverter for a load.

[0018] Figure 2 This is a schematic diagram of the data-driven control method based on an event-triggered online particle swarm algorithm according to an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the online particle swarm optimization algorithm according to an embodiment of the present invention.

[0020] Figure 4 This is a flowchart of the online particle swarm optimization algorithm according to an embodiment of the present invention.

[0021] Figure 5 This is a flowchart of the model-free predictive control according to an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] The embodiments of the present invention have Taking a three-phase two-level inverter with a load as an example, its topology is as follows: Figure 1 As shown. DC bus voltage via DC bus capacitor Input, output and a set of three phases The load is connected. The inverter consists of three bridge arms, A, B, and C, and each bridge arm contains two IGBTs with complementary control signals.

[0024] Based on Kirchhoff's Laws (KL) and introducing Clark transform to simplify calculation and control, the system model of this topology is as follows: in: yes , The shaft output voltage is determined by the state of the three-phase bridge arms. yes , Shaft output current.

[0025] After discretization using the forward Euler method, the above equation yields the following prediction model: in: and These represent the system states at the current moment and the next moment, respectively. It is the control input at the current moment; It is the sampling period.

[0026] like Figure 2 As shown, this embodiment of the invention provides a data-driven control method based on an event-triggered online particle swarm optimization algorithm, comprising the following steps: S101. Construct a data-driven model for a power electronic converter based on partial scheme dynamic linearization.

[0027] In practice, parameter mismatch or drift in the aforementioned predictive models is always unavoidable. Traditional finite control set-model predictive control methods cannot achieve the desired control effect in such cases, and may even lead to system instability. Therefore, this invention constructs a partially format dynamic linearized data-driven model to build a data-driven model, as follows: in: The change in system state within one sampling period; It is the control increment vector; Represents the depth of the data involved in the computation. >0; It is a linear parameter matrix.

[0028] in: and These represent the system states of the power electronic converter in the current and next sampling periods, respectively. It is the change value of the control quantity. and These are the control inputs for the current and previous sampling periods, respectively.

[0029] S102. An improved event-triggered online particle swarm optimization algorithm is used to estimate data-driven model parameters in real time.

[0030] The data-driven model described above is a linear model with a time-varying parameter matrix that relies solely on the system's measurements and historical data, eliminating the dependence on prior model information. Used to describe the relationship between system output and input, and estimated using appropriate methods. This allows for an accurate description of the system's dynamic characteristics. In this embodiment of the invention, an event-triggered online particle swarm optimization algorithm is designed to estimate the dynamic characteristics of the system in real time. .

[0031] Traditional particle swarm optimization (PSO) algorithms are typically used offline to solve optimization problems. In this invention, the search process of PSO is integrated with the converter's operation. Data-driven model parameters are used as particle positions in the online PSO algorithm. These particle positions (estimated model parameters) are continuously applied to the system model, and the real-time state of the power electronic converter serves as the basis for particle evaluation. The tracking errors of the power electronic converter system over multiple sampling periods are used to comprehensively evaluate the particles, thereby guiding data-driven model parameter estimation and gradually determining the system model parameters that meet control requirements. An event-triggered mechanism is also introduced: if the tracking error of the current globally optimal particle exceeds a preset threshold, the online PSO algorithm is activated; otherwise, the data model parameters remain unchanged.

[0032] The schematic diagram and flowchart of the online particle swarm optimization algorithm are as follows: Figure 3 and Figure 4 As shown, the specific steps are as follows: (1) Apply particles as data-driven model parameters.

[0033] (2) Collect the power electronic converter system state for multiple sampling periods and calculate the current particle's cost.

[0034] (3) Repeat steps (1) and (2) until all particles are evaluated.

[0035] (4) Update the current global optimum and individual optimum, and update the population position.

[0036] (5) Apply the globally optimal particle as the data-driven model parameter.

[0037] (6) Collect the power electronic converter system state for multiple sampling periods and calculate the cost of the current global optimal particle.

[0038] (7) Determine whether to continue using the online particle swarm algorithm for model parameter estimation based on the event triggering conditions; if the event triggering is successful, update the data model parameters; if the event triggering is unsuccessful, keep the data model parameters unchanged.

[0039] (8) Repeat steps (1)-(7) to adjust the transformer model parameters in real time using the online particle swarm algorithm.

[0040] In this embodiment of the invention, the parameter search dimension of the online particle swarm optimization algorithm is 2×3 ( =3), population size is Let the first The particle in the th The position, velocity, individual optimal position, and global optimal position at the next iteration are respectively , , and The cost function is designed as follows: in: in: This represents the system's expected output during the current sampling period; Represents the current state of the system; This indicates the number of sampling periods used to evaluate the particles.

[0041] The online particle swarm optimization algorithm randomly generates population positions within a given range. The particle velocity and position update methods are described as follows: in: and They are the first The updated positions of the particles; It is inertial weight; and These are individual learning factors and social learning factors.

[0042] Furthermore, since the control error of the system can never be zero in reality, the online particle swarm optimization algorithm will continuously search and update the model parameters, leading to unnecessary switching actions, reducing the converter's lifespan, and increasing maintenance costs. Therefore, an event trigger for the online particle swarm optimization algorithm is designed to control the model parameter updates, as follows: in: These are manually preset event trigger thresholds. >0; and These represent the estimated model parameters at the current and previous time points, respectively. It is particles in In the next evaluation cycle, the first The value of the next generation; It is the total value of an evaluation of a particle.

[0043] S103. Design a model-free predictive controller based on data-driven models and finite control set-model predictive control methods.

[0044] The flowchart of model-free predictive control is as follows: Figure 5 As shown, the specific process is as follows: (1) Calculate the finite control set based on the power electronic converter system.

[0045] (2) Predict the state set of the finite control set based on the current data-driven model.

[0046] (3) Calculate the cost of all control inputs based on the predicted state set.

[0047] The cost function is designed as follows: in Let be the expected current at the next moment; It is the system state at the next moment, obtained from the current system prediction model.

[0048] (4) Sort the cost values, determine the control input with the lowest cost value, and apply it as the control input to the power electronic converter.

[0049] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A data-driven control method based on an event-triggered online particle swarm optimization algorithm, characterized in that, The steps include the following: (1) Construct a data-driven model for a power electronic converter based on partial scheme dynamic linearization; (2) An improved event-triggered online particle swarm optimization algorithm is used to estimate data-driven model parameters in real time; (3) Design a model-free predictive controller based on data-driven model and finite control set-model predictive control method.

2. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 1, characterized in that, The partial scheme dynamic linearized power electronic converter data-driven model is specifically as follows: in: Represents the depth of the data involved in the computation. >0; It is the change in the system state of the power electronic converter in the next sampling period; It controls the input increment vector; It is the change in control input within a sampling period; It is the parameter matrix of the data-driven model.

3. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 1, characterized in that, The improved event-triggered online particle swarm optimization algorithm is as follows: the data-driven model parameters are used as the particle positions in the online particle swarm optimization algorithm, and the real-time system state of the power electronic converter is used as the basis for evaluating the particles; the tracking errors of the power electronic converter system in multiple sampling periods are used to comprehensively evaluate the particles, thereby guiding the estimation of data-driven model parameters; at the same time, an event-triggered mechanism is introduced, if the tracking error of the current global best particle exceeds a preset threshold, the online particle swarm optimization algorithm is started, otherwise the data model parameters remain unchanged.

4. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 3, characterized in that, The steps for updating model parameters in an event-triggered online particle swarm optimization algorithm are as follows: (2.1) Apply particles as data-driven model parameters; (2.2) Collect the power electronic converter system state for multiple sampling periods and calculate the cost of the current particle; (2.3) Repeat steps (2.1) and (2.2) until all particles have been evaluated; (2.4) Update the current global optimum and individual optimum, and update the population position; (2.5) Apply the globally optimal particle as the data-driven model parameter; (2.6) Collect the power electronic converter system state for multiple sampling periods and calculate the cost of the current global optimal particle; (2.7) Determine whether to continue using the online particle swarm optimization algorithm for model parameter estimation based on the event triggering conditions; if the event triggering condition is met, update the data model parameters; if the event triggering condition is not met, keep the data model parameters unchanged. (2.8) Repeat steps (2.1)-(2.7) to adjust the transformer model parameters in real time using the online particle swarm algorithm.

5. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 4, characterized in that, The cost function design for the event-triggered online particle swarm optimization algorithm is as follows: in: This represents the expected output of the power electronic converter system in the current sampling period; Represents the current state of the system; This indicates the number of sampling periods used to evaluate the particles.

6. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 4, characterized in that, The event-triggered mechanism of the online particle swarm optimization algorithm is as follows: in: It is the event trigger threshold. >0; and These represent the data-driven model parameter estimates for the current and previous time points, respectively. It is particles in In the next evaluation cycle, the first The value of the next generation; It is the total value of an evaluation of a particle.

7. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 1, characterized in that, Step (3) specifically involves: (3.1) Calculate the finite control set based on the power electronic converter system; (3.2) Predict the set of states of a finite control set based on the current data-driven model; (3.3) Calculate the cost of all control inputs based on the predicted state set; (3.4) Sort the cost values, determine the control input with the smallest cost value, and apply it as the control input to the power electronic converter.

8. The data-driven control method based on event-triggered online particle swarm optimization algorithm according to claim 7, characterized in that, The cost function of the model-free predictive controller is specifically as follows: in: This represents the expected output of the power electronic converter system in the next sampling period; It is the system state at the next moment obtained from the current model-free predictive controller.

Images