Multi-target model prediction control method for wind power access multi-terminal flexible direct current system

Abstract

The invention relates to a multi-target model prediction control method for a wind power access multi-terminal flexible direct current system. Comprising the following steps: constructing a grey box stochastic differential equation model based on wind power randomness and multi-terminal flexible direct current system operation data, and applying Lamperti transformation to the model to obtain an equivalent state space prediction model; establishing a multi-target optimization model considering wind power real-time tracking, DC voltage stabilization and multi-converter station collaborative power distribution in the prediction domain; setting a random stability constraint by taking moment stability as a criterion, and determining the probability constraint of a system state quantity and a control quantity into an equivalent deterministic constraint; and solving the multi-objective optimization model at each sampling moment based on the random stability constraint and the deterministic constraint, obtaining the optimal control quantity of the multiple converter stations, and applying and executing the optimal control quantity. Robust stability, real-time power tracking and multi-station cooperative distribution can be achieved under the conditions of wind power random disturbance and power electronic nonlinearity, and high control precision and real-time performance are achieved.

Description

Technical Field

[0001] This application relates to the field of power system control technology, and in particular to a multi-objective model predictive control method for wind power access to a multi-terminal flexible DC system. Background Technology

[0002] With the continuous advancement of the national energy transition strategy, wind energy, as a core component of clean and renewable energy, faces increasingly urgent demands for large-scale development and long-distance consumption. High-voltage direct current (HVDC) and multi-terminal direct current (MTDC) technologies based on voltage source converters (VSCs) have become important technical pathways for solving the problem of long-distance wind power grid connection due to their flexible control and excellent power regulation performance, providing key support for the cross-regional optimal allocation of wind power resources. However, the inherent low inertia and strong nonlinearity of MTDC networks make them highly sensitive to system power imbalances and DC voltage fluctuations, posing a severe challenge to operational stability. Especially after a high proportion of wind power is integrated, the continuous randomness of wind power output forms a complex coupling with the system control links, easily inducing broadband oscillations covering frequencies from several Hz to hundreds of Hz, further exacerbating the instability risk of the MTDC system and seriously threatening the safe and stable operation of the power grid.

[0003] To address the aforementioned issues, various control methods have been proposed in existing technologies, among which open-loop or quasi-open-loop control methods are widely used, such as droop control and DC voltage deviation control. While these methods can improve the system's ability to cope with faults to some extent, they generally suffer from insufficient adaptability to random disturbances caused by wind power, making it difficult to achieve coordinated optimization of multiple control objectives such as wind power tracking, DC voltage constraints, and power allocation among multiple converter stations, resulting in limited control effectiveness. To address the problem of insufficient system inertia, technologies such as Virtual Synchronous Generators (VSGs) have been proposed to improve the system's inertial characterization, enhancing its disturbance immunity by simulating the operating characteristics of synchronous generators. However, these methods still have significant limitations in terms of the accuracy of modeling uncertainties introduced by wind power integration and the coordination of multi-objective online optimization, failing to fully adapt to the complex requirements of high-proportion wind power scenarios. From a system modeling perspective, traditional small disturbance analysis models are based on ordinary differential equations (ODEs). These models can only reflect the dynamic characteristics of the system near a single operating point and are difficult to accurately describe the impact of stochastic continuous disturbances (SCDs) on the overall dynamic characteristics of the system. To compensate for this deficiency, some studies use simple additive / multiplicative noise modeling or methods based solely on probability density fitting (such as Gaussian or Weibull distributions) to characterize stochastic disturbances. However, these methods often deviate significantly from the actual disturbance processes in wind power integration scenarios, thus limiting the practicality of control strategies designed based on these models.

[0004] Stochastic Differential Equations (SDEs), as a modeling tool that combines the advantages of first-principles and data-driven approaches, can effectively propagate system uncertainties. Combined with the Lamperti transform, complex disturbance processes in the "random space" can be transformed into an approximate Gaussian description that is easier to linearize and use for predictive control design, providing a new approach to solving stochastic disturbance modeling problems. However, in MTDC systems with a high proportion of wind power integration, multiple weakly damped modes often coexist, and the characteristics of these modes dynamically shift with wind power fluctuations. If the control strategy designed based on SDEs only sets conventional performance penalty terms and lacks explicit constraints for stochastic stability, the cumulative effect of SDEs may still push the system away from the stable attraction region, eventually leading to chronic instability. If all eigenvalues ​​are included in probabilistic constraints to ensure stability, it will form a difficult-to-solve high-dimensional nonlinear chance constraint problem, which cannot meet the real-time requirements of the control strategy.

[0005] In summary, the current field of MTDC system control with a high proportion of wind power integration urgently needs a multi-objective model predictive control (MPC) strategy for random disturbances. This strategy must simultaneously meet two core requirements: first, it must be able to explicitly guarantee the stochastic stability of the system and suppress the instability risk caused by the accumulation of stochastic disturbances (SCD); second, it must be able to deterministically reconstruct high-dimensional probabilistic constraints to meet the real-time solution requirements of the control strategy, and ultimately achieve flexible coordination of wind power tracking and multi-station power allocation to ensure the safe and stable operation of the system. Summary of the Invention

[0006] Therefore, it is necessary to provide a multi-objective model predictive control method for wind power access to a multi-terminal flexible DC system to address the above-mentioned technical problems.

[0007] Firstly, this application provides a multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system. The method includes: A gray-box stochastic differential equation model is constructed based on the randomness of wind power and the operating data of a multi-terminal flexible DC system, and an equivalent state-space prediction model is obtained by applying the Lamperti transformation to the model. A multi-objective optimization model is established within the prediction domain, taking into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations. Stochastic stability constraints are set based on moment stability criteria, and the probabilistic constraints of system state variables and control variables are determinized into equivalent deterministic constraints. Based on the stochastic stability constraints and deterministic constraints, the multi-objective optimization model is solved at each sampling time to obtain the optimal control quantity for multiple converter stations and apply it.

[0008] Optionally, in one embodiment of this application, the gray-box stochastic differential equation is expressed as:

[0009] in, It indicates the DC bus voltage, DC current, active power of each converter station, and the status of related controls. For control including active power reference values ​​and / or droop coefficients for each terminal, The Brownian motion increment is used to characterize continuous random disturbances in wind power and the correlation between multiple wind fields. The sampling time.

[0010] Optionally, in one embodiment of this application, applying the Lamperti transformation to the model to obtain the equivalent state-space prediction model includes: The model is subjected to Lamperti transformation to constantize or equivalently eliminate the diffusion term, and linearized and discretized in the neighborhood of the running point to obtain the state update equation for the prediction domain and its mean / covariance recursion.

[0011] Optionally, in one embodiment of this application, establishing a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations includes: The overall multi-objective performance index for constructing the multi-objective optimization model is expressed as:

[0012] in, It is the overall multi-objective performance metric that needs to be minimized within the MPC framework. Let k = t, ..., t+N represent the mathematical expectation of random perturbations within the prediction range. 1 represents; Penalty for active power of receiving converter station Compared with actual wind power output The squared deviation between them; the second term is composed of Weighted power sharing targets are introduced among all VSC terminals; Indicates the first Predicted active power of each converter; , The power sharing ratio can be set to the optimal power allocation ratio; adjacency matrix elements Indicates terminal and terminal Does the power sharing error between them include it in the summation? The third term is... Weighted, DC voltage The nominal DC voltage; the fourth and fifth items are respectively determined by... and Weighting, for the system state vector and control input vector Apply a second penalty, among which and Let each of them be a weighted matrix.

[0013] Optionally, in one embodiment of this application, establishing a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations further includes: An adaptive weight scheduling and hysteresis threshold method is used to set the weights of each target and the prediction / control time domain length.

[0014] Optionally, in one embodiment of this application, the step of setting stochastic stability constraints based on moment stability as the criterion and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints includes: Impose chance constraints on state and control; The stability constraints of random small signals are characterized by a random p-order moment stability metric. The positive definite matrix is ​​obtained by solving the Lyapunov equation based on the nominal closed-loop matrix and the noise covariance, and the terminal ellipsoid is constructed as the terminal constraint.

[0015] Optionally, in one embodiment of this application, the step of setting stochastic stability constraints based on moment stability as the criterion and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints further includes: For the probabilistic conditions of voltage constraints, line current constraints, power constraints, and control quantity constraints, deterministic reconstruction is performed in the Lamperti domain using quantile tightening, moment inequalities, Chebyshev bounds, and / or conditional risk values ​​to obtain a set of deterministic constraints in linear or convex quadratic form.

[0016] Optionally, in one embodiment of this application, the step of solving the multi-objective optimization model at each sampling time based on the stochastic stability constraints and deterministic constraints to obtain the optimal control quantity for multiple converter stations and applying it includes: The rolling optimization problem is formulated as a convex QP or a QCQP with quadratic constraints; the decision variables include at least the active power reference values ​​and / or droop coefficients of each converter station. The control law adopts the structure of "offline state feedback K + online compensation quantity v". It is solved online in each sampling period, outputs the optimal control quantity and sends it to each converter station. After collecting the measurement and updating the state, it enters the next period, forming a closed-loop rolling optimization control.

[0017] The aforementioned multi-objective model predictive control method for wind power integration into multi-terminal flexible DC systems first constructs a gray-box stochastic differential equation model based on the stochasticity of wind power and the operating data of the multi-terminal flexible DC system, and applies a Lamperti transform to the model to obtain an equivalent state-space predictive model. Then, within the prediction domain, a multi-objective optimization model is established that considers real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations. Next, stochastic stability constraints are set based on moment stability as the criterion, and the probabilistic constraints of system state variables and control variables are determinized into equivalent deterministic constraints. Finally, based on the stochastic stability constraints and deterministic constraints, the multi-objective optimization model is solved at each sampling time to obtain the optimal control variables for multiple converter stations and apply them. In other words, for wind power integration into multi-terminal flexible DC (MTDC) scenarios, a stochastic multi-objective model predictive control method is proposed to address the difficulties in power tracking, DC voltage exceeding limits, and power allocation imbalance among multiple stations caused by stochastic wind power. Continuous disturbances are characterized using an Itô-type SDE grey box model, and an equivalent model that is easy to predict and handle constraints is obtained through Lamperti transformation. Within a unified optimization framework, wind power tracking, DC voltage regulation, and power distribution are simultaneously considered. Moment stability is used as a stochastic stability constraint, and probabilistic constraints such as voltage / current / power / control are determinized into linear or quadratic forms that can be solved quickly. Online rolling optimization yields the active power reference or droop parameters for each station. Compared with traditional control methods such as droop / voltage deviation or MPC that does not explicitly consider randomness, this method significantly reduces the risk of instability caused by the accumulation of random disturbances, balancing tracking accuracy, distribution fairness, and voltage safety, and possesses good real-time performance and engineering feasibility. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a multi-objective model predictive control method for wind power access to a multi-terminal flexible DC system in one embodiment. Figure 2 This is a schematic diagram of the structure of a multi-terminal DC system in one embodiment; Figure 3 This is a schematic diagram illustrating the real-time fluctuations of wind power under Gaussian and non-Gaussian distributions in one embodiment. Figure 4 This is a schematic diagram of the power distribution before and after the Lamperti transformation in one embodiment; Figure 5 This is a schematic diagram of transmission losses under different power distributions in one embodiment; Figure 6 This is a schematic diagram illustrating the effects of different probability constraint violation levels on DC voltage deviation and wind power tracking error in one embodiment. Figure 7 This is a schematic diagram illustrating the control cost in one embodiment; Figure 8This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] In one embodiment, such as Figure 1 As shown, a multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system is provided, including the following steps: S101: Based on the randomness of wind power and the operating data of the multi-terminal flexible DC system, a gray box stochastic differential equation model is constructed, and the Lamperti transformation is applied to the model to obtain an equivalent state-space prediction model.

[0021] In this embodiment of the application, the MTDC system consists of multiple VSCs and DC lines as shown in the attached figure. Figure 2 As shown, the wind farm is connected to the wind farm-side VSC via the AC side. Each VSC uses an average value model and is configured with vector control; the AC side uses... Equivalently, the DC side consists of a current source and... Parallel branch and bridge arm reactor equivalent Composition. A wind farm can be represented by an equivalent third-order wind turbine model; wind farm coordinates. With MTDC coordinates Coordinate normalization transformation and alignment to the VSC rotating coordinate system are required between them to achieve unified coupling modeling.

[0022] To characterize the continuous randomness, non-Gaussian nature, and multi-wind field correlation of wind power, an Itô process is introduced to establish a gray-box SDE model. The general form is:

[0023] in, For Wiener process increment, For drifting, The spread function is used. The mechanical power of the wind turbine exhibits a small random deviation around the predicted value.

[0024]

[0025] in, For the first The mechanical power of each fan (or equivalent unit of a fan), To predict its mechanical power at a given time, relative to the predicted value Small random bias.

[0026] By coupling the wind power SDE with the VSC-MTDC differential-algebraic equations, the overall system model can be obtained:

[0027] in, This is the system state vector that includes the wind farm and various state variables of the VSC–MTDC (such as voltage, current, power, etc.). For the corresponding control input vectors (such as active / reactive power commands for each converter station, DC voltage reference values, etc.). Let S be the vector of the wind turbine mechanical power stochastic process described by the above SDE. and These represent the system's differential equations and algebraic constraint equations, respectively.

[0028] This modeling allows for the selection of different... Characterizing non-Gaussian and correlated disturbances better reflects the wind power fluctuation characteristics at engineering sites. The wind power fluctuation characteristics obtained in the examples are shown in the attached figure. Figure 3 As shown.

[0029] In one embodiment of this application, the gray-box stochastic differential equation is expressed as:

[0030] in, It indicates the DC bus voltage, DC current, active power of each converter station, and the status of related controls. For control including active power reference values ​​and / or droop coefficients for each terminal, The Brownian motion increment is used to characterize continuous random disturbances in wind power and the correlation between multiple wind fields. The sampling time.

[0031] In one embodiment of this application, applying the Lamperti transformation to the model to obtain an equivalent state-space prediction model includes: The model is subjected to Lamperti transformation to constantize or equivalently eliminate the diffusion term, and linearized and discretized in the neighborhood of the running point to obtain the state update equation for the prediction domain and its mean / covariance recursion.

[0032] In one embodiment of this application, a Lamperti transformation is applied to the SDE to constant or equivalently eliminate the diffusion term; and linearization and discretization are performed in the neighborhood of the running point to obtain the discrete state update equation in the prediction domain, which serves as the prediction model for MPC. Wind power is modeled using a non-Gaussian process obtained through statistical identification.

[0033] To reduce the intractability of the state-dependent diffusion term and facilitate the construction of probability constraints, the Lamperti transformation is employed. New variables are defined as follows:

[0034] Select Make This normalizes the diffusion term to a constant. The transformed SDE can be obtained from the Itô formula, and based on this, the original state-dependent diffusion is transformed into state-independent diffusion, greatly simplifying the modeling and analysis of stochastic systems. The function can be derived from Determined. The continuous SDE, after discretization and a first-order Itô-Taylor expansion, yields...

[0035] The transformed variables are closer to a Gaussian distribution, which can approximate a Gaussian stochastic differential equation, thus obtaining the mean and variance, and predicting the state at future moments. Examples of implementation before and after the transformation are attached. Figure 4 As shown. For continuous-time stochastic differential equations, the discrete form can be expressed as:

[0036] According to the first-order Iton-Taylor expansion, the above equation can be approximated as:

[0037] The discrete-time version of this linear system is given by the following equation:

[0038] Based on this, the mean and covariance of the state are recursively calculated during the online phase, providing statistics for predictive control.

[0039] S102: Establish a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations.

[0040] In this embodiment of the application, a multi-objective performance index is constructed. It includes at least: tracking error terms between wind power reference and actual output, DC voltage deviation terms, deviation terms between target allocation and actual allocation for multiple converter stations, and penalty terms for state and control increments.

[0041] Based on state-space and stochastic process modeling, a multi-objective model predictive control framework for uncertain systems is designed. This framework covers the design of state feedback controllers, as well as the problem formulation, transformation, and solution process.

[0042] The predictive control input at time k consists of linear state feedback and a compensation term:

[0043] By utilizing the offline calculated state feedback matrix, the controller design is simplified, allowing the optimization problem to focus solely on the compensation amount. This reduces computational complexity by performing a solution. The system model can be rewritten as follows:

[0044] The system equations can be written for the entire prediction time domain, thus allowing offline computation throughout the entire prediction time domain. .

[0045] To design a multi-objective model predictive control (MPC) strategy that considers uncertainties, this multi-objective control involves renewable wind power tracking, power distribution among multiple converter stations, and DC voltage stabilization. Furthermore, it includes penalty terms for both state and control variables. System stability is guaranteed by probabilistic moment stability constraints. The optimization problem involves real-time updates of the power reference value and DC voltage for each converter station.

[0046] In one embodiment of this application, establishing a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations includes: The overall multi-objective performance index for constructing the multi-objective optimization model is expressed as:

[0047] in, It is the overall multi-objective performance metric that needs to be minimized within the MPC framework. Let k = t, ..., t+N represent the mathematical expectation of random perturbations within the prediction range. 1 represents; Penalty for active power of receiving converter station Compared with actual wind power output The squared deviation between them; the second term is composed of Weighted power sharing targets are introduced among all VSC terminals; Indicates the first Predicted active power of each converter; , The power sharing ratio can be set to the optimal power allocation ratio; adjacency matrix elements Indicates terminal and terminal Does the power sharing error between them include it in the summation? The third term is... Weighted, DC voltage The nominal DC voltage; the fourth and fifth items are respectively determined by... and Weighting, for the system state vector and control input vector Apply a second penalty, among which and Let each of them be a weighted matrix.

[0048] In one embodiment of this application, the establishment of a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations further includes: An adaptive weight scheduling and hysteresis threshold method is used to set the weights of each target and the prediction / control time domain length.

[0049] In one embodiment of this application, the weights of each target and the prediction / control time domain length are set according to the operating conditions, so that the model achieves a trade-off between tracking accuracy, optimal power allocation and control smoothness; the weights can be adaptively adjusted according to different control emphasis requirements.

[0050] The stage costs are weighted and summed according to four types of objectives, with non-negativity and normalization constraints applied to the weights during initialization. The weights are adaptively adjusted using a convex combination and hysteresis approach to avoid frequent fluctuations. For updating coefficients (e.g.) ), For the normalization operator, we get:

[0051]

[0052]

[0053] in, For interval projection; to suppress jitter, a double-threshold hysteresis is used for the threshold: when Forced promotion Up to the highest realm, and only when Only then is a pullback allowed. The same principle applies to tracking and allocation channels. This allows for priority voltage stabilization during sudden voltage deviations, priority tracking during sudden wind speed changes, and priority allocation during power flow congestion. The transmission losses under different power allocations in the embodiment are shown in the attached figure. Figure 5 As shown.

[0054] S103: Set stochastic stability constraints based on moment stability criteria, and convert the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints.

[0055] In this embodiment, mean-square boundedness or second-moment stability is used as the form of stochastic stability constraints to ensure that the expectation and variance of the state are bounded under continuous random disturbances. The probabilistic conditions of voltage constraints, line current constraints, power constraints, and control quantity constraints are deterministically reconstructed in the Lamperti domain using quantile tightening, moment inequalities, Chebyshev bounds, and / or conditional value of risk (CVaR) to obtain a set of deterministic constraints in linear or convex quadratic form.

[0056] Specifically, in one embodiment of this application, the step of setting stochastic stability constraints based on moment stability as the criterion, and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints, includes: S201: Impose chance constraints on state and control.

[0057] S202: The stability constraint of random small signals is characterized by a random p-order moment stability metric.

[0058] S203: Based on the nominal closed-loop matrix and noise covariance, the positive definite matrix is ​​obtained by solving the Lyapunov equation, and the terminal ellipsoid is constructed as the terminal constraint.

[0059] In one embodiment of this application, opportunity constraints are imposed on the state and control as follows:

[0060]

[0061] in, and A constant vector, and Given a constant, and These are preset constants that represent the acceptable risk level of violating state and input constraints. These chance constraints ensure that the probability of violating the specified limits of system state and control input remains below a specified threshold.

[0062] The stability of a system's p-th moment under stochastic stability conditions is characterized by a stochastic small-signal stability metric.

[0063] Where c is a nonnegative constant, if satisfying If the system is stable at its random p-th moment, then the moment stability at p=1 and p=2 is generally more important, indicating that the mean and variance of the system are bounded as t→∞ under random excitation.

[0064] The state is subject to the following terminal constraints:

[0065] Furthermore, at a given risk level The second-order moment stability can be written as a confidence ellipsoid constraint:

[0066] in, Let Jacobian matrix be the reference point. The above equation essentially gives the ellipsoidal feasible region determined by the closed-loop matrix and the noise covariance. In the closed-loop gain... When fixed, the nominal closed-loop matrix With noise terms The corresponding Lyapunov equation:

[0067] There exists a unique positive definite solution. Based on this, the terminal confidence ellipsoid can be calculated offline. As a terminal constraint, in online MPC, it is only necessary to force the terminal state to fall into the constraint. This avoids repetitive solutions at each step and improves real-time performance. The effects of different probability constraint violation levels on DC voltage deviation and wind power tracking error in the embodiments are shown in the attached figure. Figure 6 As shown. Parameters of the terminal ellipsoid. The relevant Jacobian and statistical thresholds are solidified into online parameters after offline calculation, and only the corresponding quadratic or equivalent second-order cone (SOC) constraints are applied in the online stage.

[0068] In one embodiment of this application, the step of setting stochastic stability constraints based on moment stability as a criterion and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints further includes: For the probabilistic conditions of voltage constraints, line current constraints, power constraints, and control quantity constraints, deterministic reconstruction is performed in the Lamperti domain using quantile tightening, moment inequalities, Chebyshev bounds, and / or conditional risk values ​​to obtain a set of deterministic constraints in linear or convex quadratic form.

[0069] In one embodiment of this application, since wind power disturbances generally result in a non-Gaussian distribution in the original variable domain, making direct transformation difficult, a Lamperti transformation is first applied to the existing stochastic model to obtain equivalent variables. Within the Lamperti domain, the disturbance can be approximated as a Gaussian distribution. Therefore, the state-chance constraint can be written as:

[0070] in, For the predicted time The system output vector has a mean and covariance of , respectively. and ; In order to be with the first The weight vector corresponding to each linear chance constraint. This is the upper bound scalar of the constraint; The cumulative distribution function of the standard normal distribution; This is the upper bound of the allowed probability of default under this constraint.

[0071] And it can be equivalent to a deterministic inequality:

[0072] The control quantity is obtained similarly:

[0073] To suppress the cumulative effect of random disturbances from a mechanistic perspective, this embodiment introduces a p-order moment stability constraint. For the linearized Lamperti domain model, the state increment solution can be obtained using the Itô formula:

[0074] In this embodiment, to ensure feasibility and closed-loop stability under the influence of random wind power disturbances, this step reconstructs the probability constraints of the state / control variables and the stability constraints of random small disturbances in a unified manner, resulting in deterministic inequalities and terminal invariant ellipsoids that can be solved quickly online.

[0075] S104: Based on the stochastic stability constraints and deterministic constraints, solve the multi-objective optimization model at each sampling time to obtain the optimal control quantity for multiple converter stations and apply it.

[0076] In one embodiment of this application, the step of solving the multi-objective optimization model at each sampling time based on the stochastic stability constraints and deterministic constraints to obtain the optimal control quantity for multiple converter stations and applying it includes: S301: The rolling optimization problem is formulated as a convex QP or a QCQP with quadratic constraints; the decision variables include at least the active power reference values ​​and / or droop coefficients of each converter station.

[0077] S302: The control law adopts the structure of "offline state feedback K + online compensation quantity v". It is solved online in each sampling period, outputs the optimal control quantity and sends it to each converter station. After collecting the measurement and updating the state, it enters the next period, forming a closed-loop rolling optimization control.

[0078] In one embodiment of this application, the rolling optimization problem is formulated as a convex QP or a QCQP with quadratic constraints; the decision variables include at least the active power reference value and / or droop coefficient of each converter station.

[0079] A rolling time-domain online optimization and execution mechanism is adopted. To reduce the real-time computational burden, the control law adopts a structure of "offline state feedback K + online compensation amount v":

[0080] in, For the state estimation at the current sampling time, This is the first step of online optimization to obtain the optimal compensation amount.

[0081] Based on the equivalent state-space prediction model and deterministic constraints, the nominal closed-loop prediction under compensation parameterization is:

[0082] The length is By expanding and stacking the time domain, we can obtain:

[0083] Within the unified multi-objective framework, the cost function is written as a quadratic form in the prediction domain:

[0084] Based on the obtained convex constraints, the above objective function can be expressed as a convex quadratic programming problem. To ensure convexity, the weight matrix satisfies... All quadratic constraints are either convex quadratic forms or equivalent SOC (second norm) forms.

[0085] The system solves online in each sampling cycle, outputs the optimal control quantity, and sends it to each converter station. After collecting and updating the measurement status, it enters the next cycle, forming a closed-loop rolling optimization control. The control law adopts "offline status feedback". +Online compensation amount The structure is designed to reduce the burden of real-time computing.

[0086] At sampling time Measurements (DC voltage, current, active / reactive power at each station, grid-side frequency, etc.) are collected from each converter station and DC line. Extended Kalman or unscented Kalman spectroscopy (consistent with the Lamperti domain) can be used to obtain these measurements. The optimal estimate and upper bound of covariance are used for subsequent constraint tightening and prediction.

[0087] Based on the above-mentioned forms of cohesion, with Recombined with reference trajectories (wind power reference, DC voltage reference, multi-station power target allocation coefficient) ,vector With the right-hand side of the constraint; embed the quantile terms of the chance constraint with the covariance boundary. In this context, the QP / QCQP of the current cycle is constituted.

[0088] By calling an embedded QP / QCQP solver (such as one based on the interior-point method or the primal-dual method), and using the shift vector of the previous period's solution as the initial value, the optimal compensation sequence can be obtained. .

[0089] To meet strict real-time requirements, the following acceleration strategies are supported: prefactoring. Utilization of sparse structures; fixed prediction matrix Only the right-hand side is updated; a finite iteration hot-start mode is adopted to return the current feasible optimal value within the time limit.

[0090] The final control quantity is synthesized according to the control law and mapped to the execution quantity of each converter station. Two common forms are given: active power reference type and equivalent droop parameter type. "Voltage-power" coordination is achieved by adjusting the equivalent droop. To prevent integral saturation or extreme disturbances, anti-saturation / backoff logic is set at the execution end: if the solution is infeasible or the solution timeout occurs, it temporarily backs to the previous state. Or the first step of the previous feasible solution, and resume MPC in the next cycle.

[0091] Will According to the communication protocol, the data is sent to the station control systems of each converter station and then executed to advance the time domain. Window update reference and prediction; The shift is used as the initial value for the next cycle, and failed end values ​​are cleared. The above steps are repeated to form a closed-loop rolling optimization. The control cost in the embodiment is shown in the appendix. Figure 7 As shown.

[0092] In the aforementioned multi-objective model predictive control method for wind power integration into multi-terminal flexible DC systems, firstly, a gray-box stochastic differential equation model is constructed based on the stochasticity of wind power and the operating data of the multi-terminal flexible DC system. A Lamperti transformation is then applied to this model to obtain an equivalent state-space predictive model. Next, a multi-objective optimization model is established within the prediction domain, taking into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations. Then, stochastic stability constraints are set based on moment stability as the criterion, and the probabilistic constraints of system state variables and control variables are determinized into equivalent deterministic constraints. Finally, based on the stochastic stability constraints and deterministic constraints, the multi-objective optimization model is solved at each sampling time to obtain the optimal control variables for multiple converter stations and apply them. In other words, for wind power integration into multi-terminal flexible DC (MTDC) scenarios, a stochastic multi-objective model predictive control method is proposed to address the difficulties in power tracking, DC voltage exceeding limits, and power allocation imbalance among multiple stations caused by stochastic wind power. Continuous disturbances are characterized using an Itô-type SDE grey box model, and an equivalent model that is easy to predict and handle constraints is obtained through Lamperti transformation. Within a unified optimization framework, wind power tracking, DC voltage regulation, and power distribution are simultaneously considered. Moment stability is used as a stochastic stability constraint, and probabilistic constraints such as voltage / current / power / control are determinized into linear or quadratic forms that can be solved quickly. Online rolling optimization yields the active power reference or droop parameters for each station. Compared with traditional control methods such as droop / voltage deviation or MPC that does not explicitly consider randomness, this method significantly reduces the risk of instability caused by the accumulation of random disturbances, balancing tracking accuracy, distribution fairness, and voltage safety, and possesses good real-time performance and engineering feasibility.

[0093] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0094] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0095] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0096] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0097] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0098] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system, characterized in that, The method includes: A gray-box stochastic differential equation model is constructed based on the randomness of wind power and the operating data of a multi-terminal flexible DC system, and an equivalent state-space prediction model is obtained by applying the Lamperti transformation to the model. A multi-objective optimization model is established within the prediction domain, taking into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations. Stochastic stability constraints are set based on moment stability criteria, and the probabilistic constraints of system state variables and control variables are determinized into equivalent deterministic constraints. Based on the stochastic stability constraints and deterministic constraints, the multi-objective optimization model is solved at each sampling time to obtain the optimal control quantity for multiple converter stations and apply it.

2. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The gray box stochastic differential equation is expressed as: in, It indicates the DC bus voltage, DC current, active power of each converter station, and the status of related controls. For control including active power reference values ​​and / or droop coefficients for each terminal, The Brownian motion increment is used to characterize continuous random disturbances in wind power and the correlation between multiple wind fields. The sampling time.

3. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The process of applying the Lamperti transformation to the model to obtain the equivalent state-space prediction model includes: The model is subjected to Lamperti transformation to constantize or equivalently eliminate the diffusion term, and linearized and discretized in the neighborhood of the running point to obtain the state update equation for the prediction domain and its mean / covariance recursion.

4. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The establishment of a multi-objective optimization model within the prediction domain that takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations includes: The overall multi-objective performance index for constructing the multi-objective optimization model is expressed as: in, It is the overall multi-objective performance metric that needs to be minimized within the MPC framework. Let k = t, ..., t+N represent the mathematical expectation of random perturbations within the prediction range. 1 represents; Penalty for active power of receiving converter station Compared with actual wind power output The squared deviation between them; the second term is composed of Weighted power sharing targets are introduced among all VSC terminals; Indicates the first Predicted active power of each converter; , The power sharing ratio can be set to the optimal power allocation ratio; adjacency matrix elements Indicates terminal and terminal Does the power sharing error between them include it in the summation? The third term is... Weighted, DC voltage The nominal DC voltage; the fourth and fifth items are respectively determined by... and Weighting, for the system state vector and control input vector Apply a second penalty, among which and Let each of them represent its respective weighted matrix.

5. The multi-objective model predictive control method for wind power access to a multi-terminal flexible DC system according to claim 4, characterized in that, The multi-objective optimization model established within the prediction domain, which takes into account real-time wind power tracking, DC voltage stability, and coordinated power allocation among multiple converter stations, also includes: An adaptive weight scheduling and hysteresis threshold method is used to set the weights of each target and the prediction / control time domain length.

6. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The method of setting stochastic stability constraints based on moment stability as the criterion, and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints, includes: Impose chance constraints on state and control; The stability constraints of random small signals are characterized by a random p-order moment stability metric. The positive definite matrix is ​​obtained by solving the Lyapunov equation based on the nominal closed-loop matrix and the noise covariance, and the terminal ellipsoid is constructed as the terminal constraint.

7. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The method of setting stochastic stability constraints based on moment stability as the criterion, and converting the probabilistic constraints of system state variables and control variables into equivalent deterministic constraints, further includes: For the probabilistic conditions of voltage constraints, line current constraints, power constraints, and control quantity constraints, deterministic reconstruction is performed in the Lamperti domain using quantile tightening, moment inequalities, Chebyshev bounds, and / or conditional risk values ​​to obtain a set of deterministic constraints in linear or convex quadratic form.

8. The multi-objective model predictive control method for wind power integration into a multi-terminal flexible DC system according to claim 1, characterized in that, The process of solving the multi-objective optimization model at each sampling time based on the stochastic stability constraints and deterministic constraints to obtain the optimal control quantities for multiple converter stations and applying them includes: The rolling optimization problem is formulated as a convex QP or a QCQP with quadratic constraints; the decision variables include at least the active power reference values ​​and / or droop coefficients of each converter station. The control law adopts the structure of "offline state feedback K + online compensation quantity v". It is solved online in each sampling period, outputs the optimal control quantity and sends it to each converter station. After collecting the measurement and updating the state, it enters the next period, forming a closed-loop rolling optimization control.

Images