Method and system for data-driven robust control of an SVG

Abstract

The application discloses an SVG data-driven robust control method and system, optimizes an outer ring control structure of SVG based on data-driven control optimization, so as to improve small disturbance stability of the system, designs global voltage and reactive power outer ring controllers of SVG based on a compact format model-free adaptive control structure, so as to improve dynamic response characteristics of the outer ring and improve wide-frequency oscillation suppression capacity of SVG, designs a disturbance observer based on RBF neural network to observe external disturbance and integrate into the controller, so as to further improve stable operation capacity of the system under severe working conditions. The application can be widely applied to wind power, photovoltaic and other new energy grid-connected systems containing SVG, so as to improve control performance of SVG and safe and stable operation capacity of the system.

Description

Technical Field

[0001] This invention relates to the field of stability analysis and control technology for new energy power generation systems, and in particular to an SVG data-driven robust control method and system. Background Technology

[0002] With the increasing penetration rate of new energy power generation, wind power, photovoltaic power, and other new energy sources exhibit strong volatility, intermittency, and low inertia characteristics, posing a severe challenge to the safety and stability of the power system when connected to the grid. Static Var Generators (SVG), as flexible reactive power compensation devices, play a crucial role in new energy grid-connected systems, maintaining voltage stability at the grid connection point, suppressing harmonics, and improving system dynamic response. However, in high-proportion power-electronic new energy power plants composed of SVG and new energy generation units, the complex dynamic interactions between multiple equipment control systems and between these systems and the weak grid can easily induce broadband oscillations ranging from subsynchronous to mid-to-high frequency bands. These oscillations not only degrade power quality and threaten equipment safety but may also evolve into systemic stability risks, becoming a key bottleneck restricting the efficient consumption of new energy and the safe operation of the power grid.

[0003] Traditional SVG control methods largely rely on precise mathematical models of the system. However, in practical engineering, the operating conditions of renewable energy power plants are complex and variable, equipment parameters are uncertain, and external environmental disturbances are frequent. Model-based control methods struggle to adapt to these dynamic changes, leading to decreased control performance or even system instability. Furthermore, existing traditional SVG virtual impedance control strategies lack adaptability in suppressing broadband oscillations and often only suppress oscillations in specific frequency bands, failing to address subsynchronous / supersynchronous and mid-frequency oscillations that may occur in renewable energy grid-connected systems. Therefore, there is an urgent need for an SVG control method that does not rely on precise models, possesses strong robustness, and adaptive broadband oscillation suppression capabilities. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an SVG data-driven robust control method and system to address the shortcomings of the existing technology, thereby improving the adaptability and robustness of the SVG control strategy in oscillation suppression and ensuring the safe and stable operation of new energy power plants.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an SVG data-driven robust control method, comprising the following steps:

[0006] The inner loop current command value i of the SVG is calculated using the following formula. dref (k) or i qref (k):

[0007] ;

[0008] Where u(k) corresponds to the inner loop current command value i at time k. dref (k) or i qref (k), y * (k) corresponds to the global voltage control command value v of the SVG at time k. ref Or reactive power control command value q ref y(k) corresponds to the average value v of the three-phase submodule capacitor voltage of the SVG at time k. dcm (k) or output reactive power q(k), Let be the estimated pseudo-partial derivative at time k, λ and κ be the weighting factors of the control criterion function, and ρ and γ be the step size factors. It is the output of the RBF neural network perturbation observer at time k. ;

[0009] will i dref (k) or i qref (k) Input to the current inner loop controller to generate modulation signal c a c b c c It controls the on / off switching of each submodule switching device in the SVG three-phase bridge arm.

[0010] This invention comprehensively considers tracking error, control increment, and tracking error difference, enabling SVG to obtain superior dynamic characteristics and effectively suppressing broadband oscillations from subsynchronous to mid-frequency bands that may occur in new energy power plants. It improves the adaptability and robustness of SVG control strategy in oscillation suppression, ensuring stable operation of SVG and high-proportion consumption of new energy.

[0011] The estimated pseudopartial derivative at time k The calculation formula is: ;

[0012] Where Γ is the step size factor of the pseudo-partial derivative estimation algorithm, and μ is the weight factor of the pseudo-partial derivative estimation criterion function. .

[0013] Criterion function of pseudo-partial derivative estimation algorithm The expression is:

[0014] ;

[0015] Where, φ c (k) is a pseudo-partial derivative.

[0016] This invention does not rely on the mathematical model of the controlled system. Instead, it updates the pseudo-partial derivative estimates and control law in real time using only the I / O data of the SVG (such as submodule capacitor voltage and reactive power data). When the system oscillates, it can collect and respond to system oscillation data, update the tight-form dynamic linearized data model of the controlled system by estimating the pseudo-partial derivatives in real time, and then quickly adjust the control law to cope with the oscillation. This allows the SVG output to quickly track the command, stabilize the control system, and ultimately suppress system oscillations.

[0017] The output of the RBF neural network perturbation observer at time k The expression is: ; where w(k) is the weight vector of the RBF neural network, and h(k) is the output of the neurons in the hidden layer of the RBF neural network.

[0018] This invention can estimate and compensate for unmodeled dynamics and external disturbances in the system in real time (such as the "black box" control of new energy units and the uncertainty brought about by changes in the operating conditions of new energy power plants), effectively enhancing the robustness of the control system.

[0019] The iterative formulas for the weights, hidden node widths, and hidden node center vectors of the RBF neural network are as follows: ;

[0020] Among them, h j The output of the j-th neuron in the hidden layer. Let be the weight of the j-th hidden node at time k. Let be the width of the j-th hidden node at time k. Let be the center vector of the j-th hidden node at time k. η is the learning rate in gradient descent, and α is the momentum factor. Initial weights ω j Take a random number between 0 and 1, with an initial width of b. j Let j = 1, 2, ..., 5; the initial value of the center vector is... .

[0021] Control Criterion Function The expression is:

[0022] .

[0023] As an inventive concept, the present invention also provides an SVG data-driven robust control system, which includes a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the steps of the above method.

[0024] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides an SVG data-driven robust control method. This method does not rely on a precise system mathematical model and can achieve adaptive optimization of the control strategy solely through the I / O data of the SVG. By establishing a compact-format dynamic linearized data model based on model-free adaptive control theory, it can respond to the dynamic changes of the system in real time, overcoming the limitations of oscillation suppression methods relying on renewable energy impedance models under complex operating conditions. Simultaneously, the designed RBF neural network disturbance observer can effectively observe and compensate for external disturbances and unmodeled dynamics, significantly enhancing the robustness of the control system. Furthermore, the designed control method comprehensively considers tracking error, control increment, and tracking error difference terms in the criterion function, enabling the SVG to obtain superior dynamic characteristics and effectively suppressing broadband oscillations from subsynchronous to mid-frequency bands that may occur in renewable energy power plants, ensuring the stable operation of renewable energy grid-connected systems containing SVG. This invention effectively solves the limitations of existing oscillation suppression methods based on renewable energy power plant impedance models in the "black box" scenario of renewable energy unit control, significantly improving the operational robustness of SVG and its adaptive capability for broadband oscillation suppression in complex power grid environments. Attached Figure Description

[0025] Figure 1 This is an embodiment of the new energy power station topology containing SVG according to the present invention;

[0026] Figure 2 This is a block diagram of an SVG data-driven robust control method according to an embodiment of the present invention;

[0027] Figure 3 The following are waveforms of reactive power step response under conventional control and the control proposed in this invention, representing an embodiment of the present invention.

[0028] Figure 4 The following are the current waveforms at the grid connection point of the system under conventional control and control proposed in this invention when a new energy power station experiences subsynchronous / supersynchronous oscillations: (a) shows the SVG under conventional control, and (b) shows the SVG under the SVG data-driven robust control proposed in this invention.

[0029] Figure 5 The following are waveforms of the grid-connected current of the system under conventional control and control proposed in this invention when a medium-frequency oscillation occurs at a new energy power station according to an embodiment of the present invention; (a) shows the SVG under conventional control, and (b) shows the SVG under the SVG data-driven robust control proposed in this embodiment of the present invention.

[0030] Figure 6In one embodiment of the present invention, when a medium-frequency oscillation occurs in a new energy power station, a large external data disturbance is introduced into the control. The SVG adopts the traditional model-free adaptive control without an RBF disturbance observer and the system response waveform diagram under the control proposed in this embodiment of the present invention. Detailed Implementation

[0031] 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.

[0032] Example 1

[0033] Figure 1 The topology of the new energy power station (new energy grid-connected system) with SVG in Embodiment 1 of the present invention includes direct-drive wind turbines #1-#20 and box-type transformers T1-T. 20 Internal collection lines and main transformer T m Each direct-drive wind turbine has a capacity of 1.5 MVA. The direct-drive wind turbines are connected to a 35 kV feeder via a box-type transformer. The system has four feeders, with five turbines on each feeder. A cascaded H-bridge SVG with a capacity of 10 Mvar is connected to the 35 kV bus.

[0034] like Figure 2 The diagram illustrates a block diagram of an SVG data-driven robust control (DDRC) method according to an embodiment of the present invention. This method aims to improve control robustness and enhance its adaptive suppression capability against broadband oscillations in direct-drive wind farms. In the diagram, the data-driven robust control is located in the outer loop of the SVG's global voltage and reactive power. Its specific control structure consists of a control algorithm based on the model-free adaptive control principle, a pseudo-partial derivative estimation stage, and an RBF neural network disturbance observer. The global voltage and reactive power command v... ref and q ref The corresponding instruction y in the controller * (k), the measured average voltage v of the three-phase submodule capacitor dcm The instantaneous reactive power q corresponds to the system output data y(k) in the controller, and the controller receives the control data u(k) input to the system, which corresponds to the current inner loop command i in the SVG dual closed-loop control structure. dref with i qref The embodiments of the present invention include the following steps:

[0035] 1) An equivalent linearized data model for the SVG is established based on compact dynamic linearization. The mathematical model for the instantaneous reactive power of the SVG and the mean voltage of the three-phase submodule capacitors is as follows:

[0036] ;

[0037] Where: v dcm The average value of the capacitor voltage of the SVG three-phase submodule, v dc0 v is the DC component of the capacitor voltage of a single-phase submodule. sd and i d Let d be the SVG port voltage and arm current d-axis components, q be the instantaneous reactive power output of the SVG, and i be the SVG output. q Let q be the q-axis component of the SVG bridge arm current. The instantaneous reactive power mathematical model further considers the dynamic characteristics of the inner current loop, and based on the above mathematical model, a one-step forward Euler discretization is performed to obtain:

[0038] ;

[0039] Wherein: T s v is the sampling period dcm i(k) and q(k) are the average value of the three-phase submodule capacitor voltage and reactive power output of the SVG at time k, respectively. dref (k) and i qref (k) represents the control current input at time k.

[0040] The three-phase submodule capacitor voltage average value and reactive power discrete model of the SVG in the above formula can be uniformly described as the discrete nonlinear system of SISO:

[0041] ;

[0042] Where: y(k) and u(k) are the system output and input at time k, respectively, and n y (k) and n u d(k) represents the system order, and d(k) represents the unknown bounded disturbance of the system at time k, including unknown data disturbances and unmodeled system dynamics. y(k) corresponds to the output v of the controlled system. dcm (k) or q(k), where u(k) corresponds to the input i of the controlled system. dref (k) or i qref (k). Since the data-driven robust controller setup method and structure for the d-axis and q-axis are consistent, the controller design in the following steps will uniformly use y(k), u(k), and d(k) to represent the nonlinear system. Based on model-free adaptive control theory, the compact-form dynamic linearized data model can be established as follows:

[0043] ;

[0044] Where: φ c (k) is the pseudo-partial derivative of the system, and |φ c (k)| is bounded at any time k.

[0045] 2) Design a radial basis function (RBF) neural network-based disturbance observer to observe the disturbance Δd(k) of the controlled system in real time. The output of the RBF neural network-based disturbance observer is:

[0046] ;

[0047] in: Let w(k) be the perturbation observation output at time k; w(k) = [ω1(k),...,ω n (k)] T This is the weight vector of the RBF neural network; h(k) = [h1(k),...,h...]. m (k)] T The output of the hidden layer neurons is expressed as follows:

[0048] ;

[0049] in: c is the input vector; j b is the center vector of the j-th hidden node; j Let be the width of the j-th hidden node. The iterative formulas for calculating the weights, hidden node width, and hidden node center vector using gradient descent are as follows:

[0050] ;

[0051] Where η is the learning rate in gradient descent and α is the momentum factor.

[0052] 3) Design a pseudo-partial derivative estimation algorithm to calculate pseudo-partial derivative estimates in real time. Design the following criterion function:

[0053] ;

[0054] Where μ is the weighting factor of the pseudopartial derivative estimation criterion function. Based on the above equation, the extreme values ​​of the pseudopartial derivatives are solved, and then set to zero to obtain... The real-time update algorithm is as follows:

[0055] ;

[0056] Where: Г is the step size factor of the pseudo-partial derivative estimation algorithm.

[0057] 4) Design the control algorithm u(k) for SVG data-driven robust control. Introduce an error term into the criterion function. Control increment term |Δu(k)| and tracking error difference term Therefore, the following criterion function is designed:

[0058] ;

[0059] Where: λ and Let be the weighting factor of the control criterion function. By finding the extreme value of Δu(k) and setting it to zero, while introducing a step size factor γ, the control algorithm is obtained as follows:

[0060] ;

[0061] Where: the variables in the control algorithm u(k) This is the output of the pseudo-partial derivative estimation algorithm. This is the output of the RBF neural network perturbation observer. u(k) corresponds to the SVG current inner loop command value i. dref (k) or i qref (k). y in the control algorithm * (k) corresponds to the SVG's global voltage control command value v. ref Or reactive power control command value q ref The generated i dref (k) or i qref (k) is input to the inner current loop controller to further generate the modulation signal c. a c b c c Control the on / off switching of each submodule switching device in the SVG three-phase bridge arm.

[0062] Figure 3 The following are waveforms of the reactive power step response of an SVG under conventional control and the control proposed in this embodiment of the invention. Figure 3 It can be seen that compared with traditional control, the reactive power response speed is faster and has better dynamic performance under the control proposed in the embodiments of the present invention.

[0063] Figure 4 The figures show the current waveforms at the grid connection point of a new energy power station under conventional control and the control proposed in this invention when the SVG (Static Var Generator) experiences subsynchronous / supersynchronous oscillations, according to an embodiment of this invention. In (a), the SVG uses conventional control, and in (b), the SVG uses the data-driven robust control proposed in this embodiment of the invention. Figure 3 It is known that when a new energy power station experiences subsynchronous / supersynchronous oscillations, the SVG using traditional control cannot effectively suppress the system oscillations. When the SVG adopts the data-driven robust control proposed in this invention, the oscillations are effectively suppressed, and the system operates stably.

[0064] Figure 5 The figures show the current waveforms at the grid connection point of the system under conventional control and the control proposed in this embodiment of the invention when a new energy power station experiences medium-frequency oscillation. In (a), the SVG uses conventional control, and in (b), the SVG uses the data-driven robust control proposed in this embodiment of the invention. Figure 5 It is known that when the SVG with traditional control is put into a new energy power station, the system exhibits severe oscillations. However, when the SVG with data-driven robust control proposed in this embodiment is put into the new energy power station, the system maintains stable operation without oscillations.

[0065] Figure 6 In one embodiment of the present invention, when a medium-frequency oscillation occurs in a new energy power station, a large external data disturbance is introduced into the control. The SVG employs traditional model-free adaptive control without an RBF neural network disturbance observer, and the system response waveform diagram under the control proposed in this embodiment of the present invention is shown. Figure 5 It is evident that when a large external data disturbance is introduced while the system is oscillating, the performance of traditional model-free adaptive control is affected, resulting in poor control robustness. When the data-driven robust control proposed in this embodiment is used, the system maintains stable operation, verifying its strong robustness.

[0066] Example 2

[0067] Embodiment 2 of the present invention provides a control system corresponding to Embodiment 1 above, including a memory, a processor and a computer program stored in the memory; the processor executes the computer program in the memory to implement the steps of the method of Embodiment 1 above.

[0068] In some implementations, the memory may be high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk storage device.

[0069] In other implementations, the processor can be any type of general-purpose processor, such as a central processing unit (CPU) or a digital signal processor (DSP), and there is no limitation here.

[0070] Example 3

[0071] Embodiment 3 of the present invention provides a computer-readable storage medium corresponding to Embodiment 1 above, on which a computer program / instructions are stored. When the computer program / instructions are executed by a processor, they implement the steps of the method of Embodiment 1 above.

[0072] A computer-readable storage medium can be a tangible device that holds and stores instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof.

[0073] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0074] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0077] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A robust control method driven by SVG data, characterized in that, Includes the following steps: Data-driven robust control resides in the global voltage and reactive power outer loop of the SVG. Its specific control structure consists of a control algorithm based on model-free adaptive control principles, a pseudo-partial derivative estimation stage, and an RBF neural network disturbance observer. The global voltage and reactive power commands are... v ref and q ref Commands in the corresponding controller y * ( k The measured average capacitor voltage of the three-phase submodule v dcm and instantaneous reactive power q Corresponding system output data in the controller y ( k The controller receives control data input to the system. u ( k This corresponds to the inner current loop command in the SVG dual closed-loop control structure. i dref and i qref ; The inner loop current command value i of the SVG is calculated using the following formula. dref (k) or i qref (k): ； Where u(k) corresponds to the inner loop current command value i at time k. dref (k) or i qref (k), y * (k) corresponds to the global voltage control command value v of the SVG at time k. ref Or reactive power control command value q ref y(k) corresponds to the average value v of the three-phase submodule capacitor voltage of the SVG at time k. dcm (k) or output reactive power q(k), Let be the estimated pseudo-partial derivative at time k, λ and κ be the weighting factors of the control criterion function, and ρ and γ be the step size factors. It is the output of the RBF neural network perturbation observer at time k. ; will i dref (k) or i qref (k) Input to the current inner loop controller to generate modulation signal c a c b c c Controls the on / off switching of each submodule switching device in the SVG three-phase bridge arm; The estimated pseudopartial derivative at time k The calculation formula is: ； Where Γ is the step size factor of the pseudo-partial derivative estimation algorithm, and μ is the weight factor of the pseudo-partial derivative estimation criterion function. ; Criterion function of pseudo-partial derivative estimation algorithm The expression is: ； Where, φ c (k) is a pseudo-partial derivative.

2. The SVG data-driven robust control method according to claim 1, characterized in that, The output of the RBF neural network perturbation observer at time k The expression is: ; where w(k) is the weight vector of the RBF neural network, and h(k) is the output of the neurons in the hidden layer of the RBF neural network.

3. The SVG data-driven robust control method according to claim 2, characterized in that, The iterative formulas for the weights, hidden node widths, and hidden node center vectors of the RBF neural network are as follows: ； Among them, h j The output of the j-th neuron in the hidden layer. Let be the weight of the j-th hidden node at time k. Let be the width of the j-th hidden node at time k. Let be the center vector of the j-th hidden node at time k. η is the learning rate in gradient descent, and α is the momentum factor. .

4. The SVG data-driven robust control method according to claim 1, characterized in that, Control Criterion Function The expression is: .

5. An SVG data-driven robust control system, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory; characterized in that the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 4.

Images