Port microgrid DC bus voltage intelligent control method based on virtual DC motor
By equating DC bus voltage regulation with virtual mechanical torque regulation, and combining model-free adaptive control and reinforcement learning methods, the problems of rapid response and stability of DC bus voltage control in port microgrids are solved, and the dynamic performance and steady-state accuracy under complex working conditions are improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- QINGDAO UNIV OF SCI & TECH
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-08
AI Technical Summary
Existing DC bus voltage control methods for port microgrids struggle to achieve fast and accurate voltage compensation under conditions of frequent load switching and rapid power changes. Furthermore, they are highly dependent on system model or parameter tuning, and lack robustness and adaptability.
The DC bus voltage regulation problem is equivalent to the virtual DC motor mechanical torque regulation problem. Model-free adaptive control is introduced to compensate for the virtual mechanical torque online, and reinforcement learning is combined to perform online self-tuning of the model-free adaptive controller parameters. Stable control is achieved by controlling the power converter to simulate the external characteristics of the virtual DC motor.
It improves the robustness and stability of DC bus voltage regulation, significantly enhances dynamic response performance and steady-state accuracy, and is suitable for port microgrid environments with frequent equipment start-ups and shutdowns and fluctuations in new energy output.
Smart Images
Figure CN122000855A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of port microgrid control technology, and particularly relates to an intelligent control method for DC bus voltage of port microgrid based on a virtual DC motor. Background Technology
[0002] With the continuous improvement of port electrification, intelligence, and decarbonization, microgrids, based on new energy and power electronic equipment, are increasingly widely used in port shore power systems, loading and unloading equipment power supply systems, and integrated energy systems in port areas. Microgrids offer advantages such as fewer energy conversion stages, high system efficiency, and easy integration with distributed power sources like photovoltaics and energy storage, making them an important development direction for green energy systems in ports. However, due to the strong impact and frequent fluctuations of port loads, as well as complex operating conditions, the DC bus voltage of microgrids is susceptible to load mutations, power imbalances, and parameter uncertainties, resulting in voltage fluctuations and even system stability issues. Therefore, achieving rapid, stable, and robust control of the DC bus voltage is a key technical challenge in the operation of port microgrids.
[0003] To address the issues of insufficient inertia and bus voltage fluctuations in microgrids, related research has proposed control methods based on virtual DC motors. Huang et al. proposed a virtual DC motor control strategy that introduces the mechanical and armature equations of a DC generator into the power electronic converter control, simulating the rotational inertia and damping characteristics of a DC generator. This allows the DC bus to exhibit dynamic response characteristics similar to a synchronous generator under load disturbances and fluctuations in renewable energy output, thereby improving the stability of the DC bus voltage. This type of method provides a certain energy buffering capacity for the microgrid by introducing virtual inertia and damping. However, the aforementioned virtual DC motor control methods typically use fixed virtual rotational inertia and damping parameters. Their adjustment method essentially indirectly affects the DC bus voltage by changing the system's dynamic characteristics. When the port microgrid load experiences rapid changes or severe power imbalances, relying solely on virtual inertia and damping parameters results in a certain lag in the voltage compensation process, making it difficult to effectively correct DC bus voltage deviations in a timely manner. This can easily lead to slow voltage recovery or large steady-state deviations. To improve the adaptability of virtual DC motor control methods to changes in operating conditions, Zhang et al. proposed a parameter-adaptive virtual DC generator control method. This method introduces a proportional-integral (PI) control loop into the design of the virtual moment of inertia parameters, adjusting the virtual moment of inertia online based on the DC bus voltage deviation, thereby improving the system's dynamic response speed under load disturbances. However, this method still adjusts the virtual moment of inertia parameters, and its compensation path remains an indirect adjustment method. Furthermore, the introduction of an error integral loop in the PI control can easily lead to system oscillations under conditions of frequent load switching or high system parameter uncertainty, making it difficult to meet the requirements of port microgrids for rapid and stable DC bus voltage regulation.
[0004] On the other hand, to reduce reliance on precise system models, model-free adaptive control methods have gradually gained attention. Hou Zhongsheng et al. proposed a model-free adaptive control theory. This method uses tight-form dynamic linearization to estimate the pseudo-partial derivatives of the system online using only the system input and output data, and designs the control law accordingly, achieving adaptive control without a precise system model. Model-free adaptive control exhibits good adaptability in systems with strong nonlinearity and significant model uncertainty. However, existing model-free adaptive control methods mostly treat the DC bus voltage as the direct control object, failing to effectively integrate it with the physical model of the virtual DC motor, making it difficult to target the key energy regulation quantities in the virtual DC motor. Furthermore, their controller parameters usually rely on manual experience or offline tuning, and control performance may still be affected when operating conditions change frequently.
[0005] In recent years, reinforcement learning, as an intelligent optimization method based on interaction and reward mechanisms, has gradually attracted attention in the field of control parameter self-tuning. Liu et al. attempted to introduce reinforcement learning into power electronic system control, optimizing control parameters through online learning to improve the system's adaptive capability under complex operating conditions. However, existing studies often directly apply reinforcement learning to generate control quantities, which can easily disrupt the original control structure and increase the complexity of system stability analysis and engineering implementation, thus limiting its application in microgrid control.
[0006] However, the aforementioned virtual DC motor control methods mainly rely on adjusting virtual inertia and damping parameters to indirectly affect the DC bus voltage, and the adjustment process has a certain lag. When the load undergoes rapid changes or the system power is severely unbalanced, it is difficult to compensate for the DC bus voltage deviation in a timely and effective manner through inertia and damping adjustment alone. This may still result in slow voltage recovery or large steady-state errors, and the control effect is highly sensitive to model and controller parameter settings. To reduce the dependence on precise system models, model-free adaptive control methods have gradually gained attention. This method constructs a dynamic linearized data model online using system input and output data, achieving adaptive control without precise modeling. However, model-free adaptive controllers typically contain multiple key parameters, and their control performance is highly sensitive to parameter tuning. Traditional experience or offline tuning methods are difficult to adapt to the frequently changing operating conditions and highly uncertain disturbances in port microgrids. In recent years, reinforcement learning, as an intelligent optimization method based on interaction and reward mechanisms, has shown good application potential in the field of control parameter self-tuning. However, directly using reinforcement learning to generate control commands can easily disrupt the original control structure and is not conducive to stability analysis and engineering implementation.
[0007] In summary, existing port microgrid DC bus voltage control methods primarily improve system dynamic characteristics indirectly by adjusting virtual moment of inertia or damping parameters, or directly control the bus voltage. However, these methods struggle to achieve rapid and accurate voltage compensation under conditions of frequent load switching and rapid power changes. Furthermore, these control methods are highly dependent on system models or parameter tuning, and the system's robustness and adaptability require further improvement. Therefore, it is necessary to approach the DC bus voltage regulation problem from an energy balance perspective within a virtual DC motor control framework. This involves equating the DC bus voltage regulation problem with the regulation of mechanical energy input in a virtual DC motor, introducing an adaptive control method that does not require a precise system model, performing online compensation for virtual mechanical torque, and combining this with an intelligent parameter self-tuning mechanism to improve the dynamic response performance and steady-state stability of the port microgrid DC bus voltage under complex operating conditions. Summary of the Invention
[0008] This invention addresses the issue of significant fluctuations in DC bus voltage in port microgrids under various operating conditions, including the start-up and shutdown of loading and unloading equipment, load switching, and fluctuations in renewable energy output. Existing control methods based on virtual DC motors primarily rely on virtual inertia and damping parameter adjustments, resulting in issues such as voltage compensation lag, insufficient dynamic response speed, and significant influence of parameter tuning on control performance. This invention provides a DC bus voltage control method for port microgrids based on a virtual DC motor. This method equates the DC bus voltage regulation problem to the mechanical torque regulation problem of a virtual DC motor, introduces model-free adaptive control for online compensation of the virtual mechanical torque, and further incorporates reinforcement learning for online self-tuning of the model-free adaptive controller parameters. By adaptively optimizing the control parameter configuration under different operating conditions, it effectively improves the dynamic performance and steady-state accuracy of DC bus voltage regulation, achieving rapid and stable control of the port microgrid's DC bus voltage.
[0009] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:
[0010] S1: Collect the DC bus voltage of the port microgrid, and establish and discretize the virtual DC motor model;
[0011] S2: Construct a tight-form dynamic linearized data model between virtual mechanical torque and DC bus voltage;
[0012] S3: Estimate pseudo-partial derivatives based on input and output data and design a model-free adaptive controller to achieve adaptive compensation of virtual mechanical torque;
[0013] S4: Adaptive adjustment of model-free adaptive controller parameters using reinforcement learning.
[0014] S5: By controlling the power converter to simulate the external characteristics of a DC motor, stable control of the DC bus voltage is achieved.
[0015] 1. Further, in step S1, the step of collecting the DC bus voltage of the port microgrid and establishing and discretizing the virtual DC motor model specifically includes:
[0016] (1) The output external characteristics of the simulated DC generator are realized by establishing a virtual DC motor model.
[0017] The armature equations and mechanical equations for the armature circuit of the virtual DC motor are as follows:
[0018]
[0019]
[0020] in, , Indicates the torque coefficient. Indicates magnetic flux. This is the actual angular velocity of the DC motor. The rated angular velocity of the DC motor. For rotational inertia, The damping coefficient is... For mechanical input torque, For electromagnetic torque, This refers to electromagnetic power.
[0021] (2) Discretize the armature equations by combining the mechanical equations of the virtual DC generator: The expression is as follows:
[0022]
[0023] in, This represents the number of pole pairs of the generator; Indicates time DC bus voltage; This represents the DC bus voltage at the next sampling time. Indicates time Virtual mechanical torque.
[0024] Furthermore, in step S2, the construction of a tight-form dynamic linearized data model between the virtual mechanical torque and the DC bus voltage specifically includes:
[0025] (1) Establish a discrete-time nonlinear system:
[0026]
[0027] in, Represents the system at discrete time points The system's response output at any given time. , indicating the system at discrete time points The control input, and These represent the historical lag order for the output and input variables, respectively. An unknown nonlinear function describing the dynamic characteristics of a system;
[0028] (2) Discrete-time nonlinear systems satisfy the following constraints:
[0029] The system is about and The partial derivatives exist and are continuous;
[0030] The system satisfies the generalized Lipschitz boundedness condition, that is, at any discrete sampling time t, when the input torque increment... When the value is not zero, the change in the system output satisfies: ,in , b is a positive constant.
[0031] (3) By using the tight-form dynamic linearization method, a tight-form dynamic linearization data model is established that is only related to the input virtual mechanical torque and the output DC bus voltage:
[0032]
[0033] in, for DC bus voltage at any given time; for DC bus voltage at any given time; This represents the virtual mechanical torque increment. The pseudo-partial derivative of the virtual mechanical torque with respect to the DC bus voltage represents the equivalent sensitivity of the voltage to torque changes, which is estimated online by a model-free adaptive algorithm.
[0034] Furthermore, in step S3, the step of estimating pseudo-partial derivatives based on input and output data and designing a model-free adaptive controller to achieve adaptive compensation of virtual mechanical torque specifically includes:
[0035] (1) Calculate the pseudo-partial derivative estimation law for DC bus voltage:
[0036]
[0037] in, for The estimated value, For virtual mechanical torque increment, This represents the DC bus voltage increment. Step size factor This is the weighting factor.
[0038] (2) Design a model-free adaptive controller for DC bus voltage:
[0039] Establish criterion function ;right Differentiating and setting it to zero, we obtain the control law for the virtual mechanical torque as follows:
[0040]
[0041] Where λ>0 is a weighting factor used to limit the magnitude of changes in the control input;
[0042] The desired DC bus voltage;
[0043] This is the step size factor.
[0044] Furthermore, step (1) specifically includes the following steps:
[0045] (11) The criterion function is established as follows:
[0046]
[0047] (12) Regarding both sides of the criterion function Taking the derivative and setting it to zero, we obtain the pseudo-Jacobian matrix number estimation law:
[0048]
[0049] in, This represents the pseudo-partial derivative estimation step size factor, used to adjust the update speed of the pseudo-partial derivative estimation;
[0050] This represents a regularization parameter used to prevent the denominator from approaching zero when the increment of the virtual mechanical torque is small.
[0051] Furthermore, in step S4, the adaptive adjustment of the model-free adaptive controller parameters using reinforcement learning specifically includes:
[0052] (1) Construct a reinforcement learning state vector, using the current parameters of the model-free adaptive controller as the reinforcement learning state input. The state vector is represented as follows:
[0053]
[0054] in, To control the gain weighting parameters, To control the regularization parameters, To estimate the canonical parameters using pseudo-partial derivatives, The step size parameter is estimated using pseudo-partial derivatives;
[0055] (2) Based on the state vector, construct the reinforcement learning action space and adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Perform joint adjustments to update the parameters as follows:
[0056]
[0057]
[0058]
[0059]
[0060] (3) Construct a reinforcement learning reward function, which is related to the DC bus voltage tracking error and the change in control input, and is defined as follows:
[0061]
[0062] in, This is the reference value for the DC bus voltage. This is the actual DC bus voltage. For virtual mechanical torque increment, This is the penalty coefficient.
[0063] (4) Construct a discount reward function based on the reward function:
[0064]
[0065] in, Indicates from time The initial discount reward value; Indicates time The instant reward function value; Indicates the first Discount weighting of returns at each future moment.
[0066] (5) Update the reinforcement learning policy based on the discount reward function, and, under the premise of satisfying the parameter constraints, adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Adaptive adjustment optimizes the dynamic and steady-state control performance of the DC bus voltage while ensuring system stability.
[0067] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0068] (1) Compared with the traditional control method that takes DC bus voltage as the direct control object, this invention is based on a virtual DC motor model, which equates the DC bus voltage regulation problem to a virtual mechanical torque regulation problem. By adjusting the virtual mechanical torque, the DC bus voltage deviation is compensated, avoiding the dependence on precise system models and parameters in direct voltage control, and improving the robustness and stability of DC bus voltage regulation.
[0069] (2) Compared with the virtual DC motor control method that relies solely on virtual inertia and damping parameter adjustment, this invention introduces a model-free adaptive controller to compensate for virtual mechanical torque online. It can quickly respond to DC bus voltage deviation under conditions such as load change and power imbalance, overcome the compensation lag problem of traditional virtual inertia and damping adjustment, and significantly improve the dynamic response performance of DC bus voltage regulation.
[0070] (3) Compared with modelless control methods that rely on human experience or offline parameter tuning, this invention further introduces reinforcement learning to perform online self-tuning of the parameters of the modelless adaptive controller, so that the control parameters can be adaptively optimized according to the system operating state and voltage regulation performance, thereby taking into account both the dynamic performance and steady-state accuracy of the system under different operating conditions, and enhancing the adaptive capability and robustness of the control system.
[0071] (4) The method of the present invention has a clear structure and a simple implementation process. It only needs to simulate the external characteristics of the virtual DC motor by controlling the power converter and compensate the virtual mechanical torque to achieve stable regulation of the DC bus voltage. It is suitable for the port microgrid operation environment where loading and unloading equipment frequently starts and stops and the output of new energy fluctuates significantly, and has good engineering application value. Attached Figure Description
[0072] Figure 1 This is a flowchart of an embodiment of the intelligent control method for DC bus voltage of a port microgrid based on a virtual DC motor proposed in this invention;
[0073] Figure 2 To improve the schematic diagram of the virtual DC motor;
[0074] Figure 3 This is a block diagram of bus voltage control for a port DC microgrid according to the present invention. Detailed Implementation
[0075] To make the objectives, technical solutions, and beneficial effects of the embodiments of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings of the embodiments of the present invention.
[0076] This invention addresses the issues of DC bus voltage fluctuations in port microgrids under operating conditions such as frequent start-up and shutdown of loading and unloading equipment, load switching, and fluctuations in renewable energy output. Furthermore, existing control methods based on virtual DC motors primarily rely on virtual inertia and damping parameter adjustment, lag compensation, and difficulties in adaptive tuning of control parameters. This invention proposes a DC bus voltage control method for port microgrids based on a virtual DC motor. An embodiment of this DC bus voltage control method is described in detail below.
[0077] Please see Figures 1-3 This invention proposes a method for controlling the DC bus voltage of a port microgrid based on a virtual DC motor. The method first acquires the DC bus voltage signal of the port microgrid and establishes and discretizes a virtual DC motor model. Based on this, a tight-form dynamic linearized data model is constructed between the virtual mechanical torque and the DC bus voltage. Further, the pseudo-partial derivatives of the virtual mechanical torque with respect to the DC bus voltage are calculated based on the system input and output data, and a model-free adaptive controller is designed accordingly to achieve online adaptive compensation of the virtual mechanical torque. Simultaneously, reinforcement learning is introduced to adaptively adjust the parameters of the model-free adaptive controller, enabling the control parameters to be dynamically optimized according to changes in operating conditions. Finally, by controlling the power converter to simulate the external characteristics of the virtual DC motor, stable control of the DC bus voltage of the port microgrid is achieved.
[0078] Figure 1 This is a flowchart illustrating an embodiment of a port microgrid DC bus voltage control method based on a virtual DC motor, which specifically includes the following steps:
[0079] Step 1: Collect the DC bus voltage of the port microgrid and establish and discretize the virtual DC motor model:
[0080] (1) The output external characteristics of the simulated DC generator are achieved by establishing a virtual DC motor model:
[0081] The armature equations and mechanical equations for the armature circuit of the virtual DC motor are as follows:
[0082] (1)
[0083] (2)
[0084] in, , Indicates the torque coefficient. Indicates magnetic flux. This is the actual angular velocity of the DC motor. The rated angular velocity of the DC motor. For rotational inertia, The damping coefficient is... For mechanical input torque, For electromagnetic torque, This refers to electromagnetic power.
[0085] (2) Discretize the armature equations by combining the mechanical equations of the virtual DC generator: The expression is as follows:
[0086] (3)
[0087] in, The number of pole pairs of the generator. Indicates time DC bus voltage, This indicates the DC bus voltage at the next sampling time. Indicates time Virtual mechanical torque.
[0088] Step 2: Construct a tight-form dynamic linearized data model of the relationship between virtual mechanical torque and DC bus voltage:
[0089] (1) Establish a discrete-time nonlinear system:
[0090] (4)
[0091] in, Represents the system at discrete time points The system's response output at any given time. , indicating the system at discrete time points The control input, and These represent the historical lag order for the output and input variables, respectively. It is an unknown nonlinear function that describes the dynamic characteristics of the system.
[0092] (2) The discrete-time nonlinear system involved in step (1) satisfies the following constraints:
[0093] The system is about and The partial derivatives exist and are continuous;
[0094] The system satisfies the generalized Lipschitz boundedness condition, that is, at any discrete sampling time... When the input torque increment When the value is not zero, the change in the system output satisfies: ,in , b is a positive constant.
[0095] (3) By using the tight-form dynamic linearization method, a tight-form dynamic linearization data model is established that is only related to the input virtual mechanical torque and the output DC bus voltage:
[0096] (5)
[0097] in, for DC bus voltage at any time for DC bus voltage at any time For virtual mechanical torque increment, The pseudo-partial derivative of the virtual mechanical torque with respect to the DC bus voltage represents the equivalent sensitivity of the voltage to torque changes, which is estimated online by a model-free adaptive algorithm.
[0098] Step 3: Estimate pseudo-partial derivatives based on input and output data and design a model-free adaptive controller to achieve adaptive compensation of virtual mechanical torque. This specifically includes:
[0099] (1) Calculate the pseudo-partial derivative estimation law for DC bus voltage:
[0100] (11) The criterion function is established as follows:
[0101] (6)
[0102] (12) Regarding both sides of the criterion function Taking the derivative and setting it to zero, we obtain the pseudo-Jacobian matrix number estimation law:
[0103] (7)
[0104] in, for The estimated value; This represents the virtual mechanical torque increment. This represents the DC bus voltage increment. Step size factor; This is the weighting factor.
[0105] (2) Design a model-free adaptive controller for DC bus voltage:
[0106] Establish criterion function ;right Differentiating and setting it to zero, we obtain the control law for the virtual mechanical torque as follows:
[0107] (8)
[0108] Where λ>0 is a weighting factor used to limit the magnitude of changes in the control input;
[0109] The desired DC bus voltage;
[0110] This is the step size factor.
[0111] Step 4: Adaptively adjust the parameters of the model-free adaptive controller using reinforcement learning, specifically including:
[0112] (1) Construct a reinforcement learning state vector, using the current parameters of the model-free adaptive controller as the reinforcement learning state input. The state vector is represented as follows:
[0113] (9)
[0114] in, To control the gain weighting parameters, To control the regularization parameters, To estimate the canonical parameters using pseudo-partial derivatives, The step size parameter is estimated using pseudo-partial derivatives.
[0115] (2) Based on the state vector, construct the reinforcement learning action space and adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Perform joint adjustments to update the parameters as follows:
[0116] (10)
[0117] (11)
[0118] (12)
[0119] (13)
[0120] (3) Construct a reinforcement learning reward function, which is related to the DC bus voltage tracking error and the change in control input, and is defined as follows:
[0121] (14)
[0122] in, Indicates time The immediate reward value for reinforcement learning is used to evaluate the DC bus voltage regulation performance under the current control parameter configuration; This is the reference value for the DC bus voltage; This is the actual DC bus voltage; This represents the virtual mechanical torque increment. This is the penalty coefficient.
[0123] (4) Construct a discount reward function based on the reward function:
[0124] (15)
[0125] in, Indicates from time The initial discount return value is used to comprehensively evaluate the overall impact of the current control parameter configuration on the system control performance over a future period of time; Indicates time The instantaneous return function value is used to reflect the DC bus voltage tracking performance and the smoothness of virtual mechanical torque regulation at that moment. Indicates the first The discount weight of the return at each future moment, as... As the value increases, its impact on cumulative returns gradually decreases.
[0126] (5) Update the reinforcement learning policy based on the discount reward function, and, under the premise of satisfying the parameter constraints, adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Adaptive adjustment optimizes the dynamic and steady-state control performance of the DC bus voltage while ensuring system stability.
[0127] Step 5: By controlling the power converter to simulate the external characteristics of a virtual DC motor, stable control of the DC bus voltage of the port microgrid is achieved.
[0128] This embodiment presents a port microgrid DC bus voltage control method based on a virtual DC motor. By introducing a virtual DC motor model, the DC bus voltage regulation problem is equivalent to a virtual mechanical torque regulation problem. Based on the collected port microgrid DC bus voltage signal, a dynamic model of the virtual DC motor is established and discretized. A compact-format dynamic linearization technique is used to construct an equivalent data model of the unknown port microgrid, which is only related to the input virtual mechanical torque and the output DC bus voltage. On this basis, a model-free adaptive control method based on compact-format dynamic linearization is used to calculate the virtual mechanical torque compensation, and a pseudo-partial derivative estimation algorithm is introduced to estimate the dynamic relationship between the virtual mechanical torque and the DC bus voltage online. Furthermore, based on the model-free adaptive control framework, a reinforcement learning parameter tuning mechanism is constructed to adaptively adjust the control gain weight parameters, regularization parameters, and pseudo-partial derivative estimation parameters online. Finally, the power converter control loop simulates the external characteristics of the virtual DC motor, outputting equivalent electromagnetic power to achieve stable tracking of the DC bus voltage to the reference value.
[0129] Therefore, the DC bus voltage control method described in this embodiment achieves DC bus voltage regulation through model-free adaptive control and utilizes reinforcement learning to tune the parameters of the model-free adaptive controller online, thereby stabilizing the DC bus voltage in the port microgrid. This solves the problem that traditional control methods struggle to balance dynamic response speed and steady-state accuracy under conditions of frequent start-up and shutdown of loading and unloading equipment, fluctuations in renewable energy output, and various uncertain disturbances. This method can effectively suppress DC bus voltage fluctuations under complex operating conditions, improve the real-time performance and stability of virtual mechanical torque compensation, meet the operational control requirements of the port microgrid under multi-disturbance and parameter uncertainty environments, and is applicable to port microgrid systems with different structures and capacity configurations, demonstrating good engineering applicability and scalability.
[0130] The above content describes the technical concept of this invention. Those skilled in the art can make various corresponding changes, modifications, simplifications, and combinations based on the technical solutions and concepts described above, and all such changes, modifications, simplifications, and combinations are included within the protection scope of the claims of this invention.
Claims
1. A method for intelligent control of DC bus voltage in a port microgrid based on a virtual DC motor, characterized in that, The method includes the following steps: S1: Collect the DC bus voltage of the port microgrid, and establish and discretize the virtual DC motor model; S2: Construct a tight-form dynamic linearized data model between virtual mechanical torque and DC bus voltage; S3: Estimate pseudo-partial derivatives based on input and output data and design a model-free adaptive controller to achieve adaptive compensation of virtual mechanical torque; S4: Adaptively adjust the parameters of the model-free adaptive controller using reinforcement learning; S5: By controlling the power converter to simulate the external characteristics of a DC motor, stable control of the DC bus voltage is achieved.
2. The method for controlling the DC bus voltage of a port microgrid based on a virtual DC motor according to claim 1, characterized in that: In step S1, the specific steps of collecting the DC bus voltage of the port microgrid and establishing and discretizing the virtual DC motor model include: (1) The output external characteristics of the simulated DC generator are achieved by establishing a virtual DC motor model: The armature equations and mechanical equations for the armature circuit of the virtual DC motor are as follows: in, ; Indicates the torque coefficient; Indicates magnetic flux; This represents the actual angular velocity of the DC motor. This refers to the rated angular velocity of the DC motor. It is the moment of inertia; The damping coefficient; For mechanical torque; Electromagnetic torque; Electromagnetic power; (2) Discretize the armature equations by combining the mechanical equations of the virtual DC generator: The expression is as follows: in, This represents the number of pole pairs of the generator; Indicates time DC bus voltage; This represents the DC bus voltage at the next sampling time. express Virtual mechanical torque at any given moment.
3. The method for controlling the DC bus voltage of a port microgrid based on a virtual DC motor according to claim 1, characterized in that: In step S2, the construction of a tight-form dynamic linearized data model between the virtual mechanical torque and the DC bus voltage specifically includes: (1) Establish a discrete-time nonlinear system: in, Represents the system at discrete time points The system's response output at a given time; , indicating the system at discrete time points Control input; and These represent the historical lag order for the output and input variables, respectively. An unknown nonlinear function describing the dynamic characteristics of a system; (2) Discrete-time nonlinear systems satisfy the following constraints: The system is about and The partial derivatives exist and are continuous; The system satisfies the generalized Lipschitz boundedness condition, that is, at any discrete sampling time t, when the input torque increment... When the value is not zero, the change in the system output satisfies: ,in , b is a positive constant; (3) By using the tight-form dynamic linearization method, a tight-form dynamic linearization data model is established that is only related to the input virtual mechanical torque and the output DC bus voltage: in, for DC bus voltage at any given time; for DC bus voltage at any given time; This represents the virtual mechanical torque increment. The pseudo-partial derivative of the virtual mechanical torque with respect to the DC bus voltage represents the equivalent sensitivity of the voltage to torque changes, which is estimated online by a model-free adaptive algorithm.
4. The method according to claim 1, characterized in that: In step S3, the step of estimating pseudo-partial derivatives based on input and output data and designing a model-free adaptive controller to achieve adaptive compensation of virtual mechanical torque specifically includes: (1) Calculate the pseudo-partial derivative estimation law for DC bus voltage: in, for The estimated value; This represents the virtual mechanical torque increment. This represents the DC bus voltage increment. Step size factor; As a weighting factor; (2) Design a model-free adaptive controller for DC bus voltage: Establish criterion function ;right Differentiating and setting it to zero, we obtain the control law for the virtual mechanical torque as follows: Where λ>0 is a weighting factor used to limit the magnitude of changes in the control input; The desired DC bus voltage; This is the step size factor.
5. The method according to claim 4, characterized in that: Step (1) specifically includes the following steps: (11) The criterion function is established as follows: (12) Regarding both sides of the criterion function Taking the derivative and setting it to zero, we obtain the pseudo-Jacobian matrix number estimation law: in, This represents the pseudo-partial derivative estimation step size factor, used to adjust the update speed of the pseudo-partial derivative estimation; This represents a regularization parameter used to prevent the denominator from approaching zero when the increment of the virtual mechanical torque is small.
6. The method according to claim 1, characterized in that: In step S4, the adaptive adjustment of the model-free adaptive controller parameters using reinforcement learning specifically includes: (1) Construct a reinforcement learning state vector, using the current parameters of the model-free adaptive controller as the reinforcement learning state input. The state vector is represented as follows: in, To control the gain weighting parameters; To control the regularization parameters; Estimating the canonical parameters using pseudo-partial derivatives; The step size parameter is estimated using pseudo-partial derivatives; (2) Based on the state vector, construct the reinforcement learning action space and adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Perform joint adjustments to update the parameters as follows: (3) Construct a reinforcement learning reward function, which is related to the DC bus voltage tracking error and the change in control input, and is defined as follows: in, This is the reference value for the DC bus voltage; This is the actual DC bus voltage; This represents the virtual mechanical torque increment. This is the penalty coefficient; (4) Construct a discount reward function based on the reward function: in, Indicates from time The initial discount reward value; Indicates time The instant reward function value; Indicates the first Discount weighting of returns at each future moment; (5) Update the reinforcement learning policy based on the discount reward function, and, under the premise of satisfying the parameter constraints, adjust the control gain weight parameters. Controlling regularization parameters Pseudo-partial derivative estimation of canonical parameters And pseudo-partial derivative estimation step size parameters Adaptive adjustment optimizes the dynamic and steady-state control performance of the DC bus voltage while ensuring system stability.