A wind storage safety scheduling method and related device

Abstract

The present application belongs to the technical field of offshore wind farm energy storage system, and discloses a wind storage safety scheduling method and related device. First, a discrete linear nominal dynamic model of the wind storage system is constructed, and a bounded additive disturbance set that dynamically changes with the prediction distance is introduced. By establishing the propagation mapping relationship from disturbance to state, the feedback gain matrix and the minimum robust positive invariant set are solved offline to obtain the reference safety radius of the tube section. Then, the tube section radius of each prediction period is adaptively calculated according to the time-varying disturbance boundary, the original constraint is inwardly tightened to generate a time-varying tightening constraint set, and a nominal MPC model is constructed with the measured state as the starting point to solve the optimal nominal control sequence. Finally, the actual command is corrected through linear state feedback. The present application can guarantee the absolute safety of the energy storage system hard constraint under extreme weather conditions, significantly improve the annual scheduling economy, perfectly balance the safety and economy, and effectively meet the engineering application requirements of offshore wind farm energy storage system.

Description

Technical Field

[0001] This invention belongs to the technical field of offshore wind farm energy storage systems, and particularly relates to a wind-storage safe dispatching method and related devices. Background Technology

[0002] With the rapid development of the new energy industry, offshore wind power has become an important component of the clean energy supply for the power system. However, due to extreme weather conditions such as typhoons and severe convection, offshore wind power output can fluctuate drastically in a short period, easily triggering abnormal operating conditions such as exceeding the state of charge (SOC) limit of the supporting energy storage system and backfeeding to the grid, leading to equipment damage and penalties in dispatch assessments. In these safety-critical operating scenarios, the dispatch control algorithm must prioritize the absolute guarantee of hard constraints; its performance directly determines the operational safety and grid connection stability of the offshore wind farm's energy storage system.

[0003] Currently, most offshore wind farm energy storage system scheduling schemes adopt model predictive control architecture. The mainstream technical routes can be divided into three categories, all of which have technical defects: The first category is the fixed deterministic MPC scheme, which optimizes scheduling based solely on single-point wind power prediction results, without reserving operational safety margins. Furthermore, the rolling correction mechanism has inherent lag, and sudden wind power fluctuations within the scheduling cycle can easily cause energy storage SOC to exceed limits, resulting in a high proportion of constraint violation periods throughout the year. The second category is the soft constraint penalty MPC scheme, which indirectly suppresses constraint violations by setting high penalties in the optimization objective function. This only achieves statistical constraint protection and lacks mathematically deterministic safety guarantees. In extreme electricity price scenarios, the optimizer will actively choose to exceed limits to obtain economic benefits. At the same time, the penalty coefficient needs to be manually adjusted repeatedly, resulting in poor engineering versatility. The third category is the classic minimax robust MPC scheme, which uniformly adopts the maximum disturbance interval across the entire time domain to tighten constraint boundaries, maintaining equal conservatism throughout the entire time period. However, it fails to match the objective law that wind power error amplifies with the square root of the prediction distance, which excessively compresses the energy storage adjustment space in the near-term period, leading to significant economic losses for the system throughout the year.

[0004] It is evident that existing MPC scheduling schemes cannot simultaneously ensure deterministic security under extreme weather conditions and maintain economic efficiency in scheduling operations, thus failing to meet the practical engineering application needs of offshore wind farm energy storage systems. Summary of the Invention

[0005] This invention provides a wind power storage safety dispatching method and related devices. This method can effectively solve the problem that existing MPC dispatching schemes cannot simultaneously take into account the deterministic safety guarantee of hard constraints under extreme weather conditions and the economic efficiency of dispatching operation, and can meet the actual engineering application needs of offshore wind farm energy storage systems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A wind-storage safety dispatching method includes: A discrete linear nominal dynamic model of the wind-storage system is constructed. A bounded additive disturbance set that dynamically changes with the prediction distance is constructed for the wind power prediction deviation. The propagation mapping relationship from wind power disturbance to system state is established, and the system modeling parameters and time-varying disturbance boundary containing the nominal system state equation are obtained. Based on the discrete linear nominal dynamic model of the wind storage system, the deviation dynamic equations of state deviation and control deviation are derived. The feedback gain matrix of the auxiliary linear state feedback controller is solved offline by combining the system modeling parameters. The minimum robust positive invariant set is calculated offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. Combining the reference safety radius parameter of the tube section and the time-varying disturbance boundary, the time-varying tube section radius corresponding to each prediction period is adaptively calculated based on the prediction distance, and the original hard constraints of the system are tightened inward based on the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. Taking the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, the nominal system dynamic constraints are constructed based on the discrete linear nominal dynamic model. The nominal MPC optimization model is built by combining the time-varying tightened constraint set. The optimal nominal control sequence and corresponding nominal state trajectory are obtained by solving the nominal MPC optimization model. The nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period. The feedback gain matrix is ​​invoked, and the real-time state deviation is calculated by combining the real-time collected system state and nominal state trajectory. The current basic scheduling instructions are corrected through linear state feedback, and the actual execution control instructions are generated and issued for execution.

[0007] Furthermore, the construction of the discrete linear nominal dynamic model of the wind-storage system includes: The system state vector is formed by selecting the energy storage state of charge and the power at the grid connection point of the power station; the control vector is formed by selecting the energy storage charging power and discharging power; and the exogenous input vector is formed by selecting the wind power forecast power and the day-ahead market electricity price. The discrete-time linear nominal system state equation is established to obtain the discrete linear nominal dynamic model of the wind-storage system. The bounded additive perturbation set is constructed based on the normal distribution statistical characteristics of wind power prediction errors, and the width of the perturbation interval increases with the square root of the prediction step. The propagation mapping relationship is used to characterize the integral cumulative effect of wind power disturbance in the state of charge of energy storage, as well as the direct superposition effect of wind power disturbance on the power at the grid connection point.

[0008] Furthermore, the process of solving the reference safety radius parameter of the pipe cross-section includes: By subtracting the actual system from the nominal system to eliminate deterministic input terms, the linear deviation dynamic equation is obtained. The feedback gain matrix is ​​solved; the optimal linear quadratic feedback gain matrix is ​​obtained by solving the steady-state solution of the discrete algebraic Riccati equation, so that the closed-loop deviation dynamic system satisfies discrete asymptotic stability. The maximum perturbation set in the entire prediction time domain is selected as the calculation benchmark. The minimum robust positive invariant set is calculated by the Minkowski summation iterative approximation method. The support function values ​​of each dimension of the convex polyhedron set obtained after iterative convergence are used as the benchmark safety radius parameter of the pipe section.

[0009] Further, the adaptive calculation of the time-varying tube cross-sectional radius corresponding to each prediction period based on the prediction distance includes: Linear interpolation is used for calculation, so that the near-end prediction period corresponds to a smaller pipe cross-section radius, and the far-end prediction period corresponds to a larger pipe cross-section radius that is close to the pipe cross-section reference safety radius; The original hard constraints of the system are tightened inward. Specifically, the constraint tightening amount is calculated based on the time-varying tube cross-sectional radius and the constraint function gradient. The original state constraints are shrunk into the feasible region to obtain the time-varying tightened constraint set corresponding to each prediction period, forming an adaptive constraint adjustment mechanism that is loose in the near term and tight in the far term.

[0010] Furthermore, the nominal MPC optimization model takes minimizing the comprehensive operating cost within the finite scheduling time domain as its optimization objective. The comprehensive operating cost includes the net cost of electricity purchase and sale, the cost of energy storage operation attenuation, and the cost of power deviation assessment. The nominal MPC optimization model is also equipped with inherent constraints on charging and discharging power and robust terminal invariant set constraints. After solving for the optimal nominal control sequence, the nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period.

[0011] Furthermore, each round of nominal MPC optimization forces the current measured actual operating state of the wind-storage system as the starting point of the nominal trajectory, eliminating the initial state deviation of a single round of optimization; within a single scheduling cycle, the energy storage charging and discharging power is continuously corrected through linear state feedback to suppress the continuous expansion of the state deviation caused by wind power disturbances along the energy storage charge state integral channel; after each round of scheduling ends, the starting point of the next round of nominal trajectory is reset with the latest measured actual system state, and the accumulated state deviation of the previous cycle is cleared.

[0012] Furthermore, at the end of each scheduling cycle, the actual wind power output and energy storage state of charge data are collected to calculate the state deviation at the end of the cycle; When the state deviation at the end of the cycle is detected to exceed the cross-sectional radius of the time-varying tube in the corresponding time period, it is determined that the tube boundary violation warning is triggered, the upper limit of the global disturbance set is amplified according to the preset ratio, and the cross-sectional radius of the time-varying tube and the time-varying tightening constraint set of each time period are updated simultaneously. When no boundary violation warning occurs for several consecutive scheduling cycles, the upper limit of the disturbance set is gradually reduced to the baseline value calculated offline according to a preset ratio, and the scheduling process is executed cyclically to achieve full-cycle rolling optimization.

[0013] A wind-storage safety dispatching system, comprising: The model building module is used to construct a discrete linear nominal dynamic model of the wind-storage system, construct a bounded additive disturbance set that dynamically changes with the prediction distance for wind power prediction deviation, and establish the propagation mapping relationship from wind power disturbance to system state, thereby obtaining system modeling parameters and time-varying disturbance boundaries that include the nominal system state equation. The derivation and solution module is used to derive the deviation dynamic equations of state deviation and control deviation based on the discrete linear nominal dynamic model of the wind storage system, solve the feedback gain matrix of the auxiliary linear state feedback controller offline in combination with the system modeling parameters, and calculate the minimum robust positive invariant set offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. The calculation module is used to combine the reference safety radius parameter of the tube section and the time-varying disturbance boundary, adaptively calculate the time-varying tube section radius corresponding to each prediction period according to the prediction distance, and perform inward tightening processing on the original hard constraints of the system according to the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. The instruction extraction module is used to take the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, construct the nominal system dynamic constraints based on the discrete linear nominal dynamic model, build the nominal MPC optimization model in combination with the time-varying tightened constraint set, solve the nominal MPC optimization model to obtain the optimal nominal control sequence and the corresponding nominal state trajectory, and extract the nominal control quantity of the first time period as the basic scheduling instruction for the current period. The feedback correction module is used to call the feedback gain matrix, calculate the real-time state deviation by combining the real-time collected system state and nominal state trajectory, correct the current basic scheduling instructions through linear state feedback, generate actual execution control instructions and issue them for execution.

[0014] A wind-storage safety dispatching device, comprising: Memory, used to store computer programs; A processor is used to implement the above-described wind and energy storage safety scheduling method when executing the computer program.

[0015] A computer-readable storage medium storing a computer program, which, when executed by a processor, is used to implement the above-described wind-storage safety scheduling method.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a wind power and energy storage safety scheduling method. First, a discrete linear nominal dynamic model of the wind power and energy storage system is constructed, and a bounded additive disturbance set that dynamically changes with the prediction distance is introduced. By establishing a propagation mapping relationship from disturbance to state, the feedback gain matrix and the minimum robust positive invariant set are solved offline to obtain the baseline safe radius of the pipe section. Then, based on the time-varying disturbance boundary, the pipe section radius for each prediction period is adaptively calculated, and the original constraints are tightened inward to generate a time-varying tightened constraint set. Starting from the measured state, a nominal MPC model is constructed to solve for the optimal nominal control sequence. Finally, the actual command is corrected through linear state feedback. This method utilizes the time-varying disturbance set to match the objective law that wind power prediction error increases with prediction distance, avoiding the excessive conservatism of the maximum disturbance boundary in the entire time domain of traditional robust MPC. Simultaneously, the offline robust positive invariant set ensures that the actual state is always constrained within the pipe under the worst-case disturbance, thus providing a deterministic safety guarantee in a mathematical sense. The nominal MPC only optimizes the nominal state, significantly reducing the online computational burden, while the feedback correction compensates for deviations in real time. This method ensures the absolute safety of the energy storage system under extreme weather conditions while significantly improving the economic efficiency of annual dispatch, perfectly balancing safety and economy, and effectively meeting the engineering application needs of offshore wind farm energy storage systems. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the implementation of a wind-storage safety scheduling method provided in this embodiment of the invention; Figure 2 This is a main flowchart of a wind and energy storage safety scheduling method provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a wind storage safety dispatching system provided in an embodiment of the present invention. Detailed Implementation

[0018] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0019] The technical terms involved in this invention are explained as follows: Tube-based Robust Model Predictive Control (Tube-MPC): A branch of Robust MPC, it constrains the state trajectory of the disturbed system within a tube centered on the nominal trajectory. The controller is decomposed into two parts: the nominal MPC and the auxiliary feedback controller. The cross-sectional radius of the tube ensures that all possible trajectories within the tube strictly satisfy the constraints. This invention applies it to wind and energy storage scheduling, treating wind power prediction errors as additive disturbances to construct a safe tube for energy storage SOC.

[0020] Minimal Robust Positively Invariant Set (mRPI): For a disturbed dynamic system, if the initial state is within the set, then for any permissible perturbation sequence, the system state always remains within the set; therefore, this set is a robust positively invariant set. mRPI is a minimal set containing the origin and satisfying the above property, used to determine the minimum pipe cross-sectional radius required to guarantee constraint satisfaction in pipe MPC.

[0021] Nominal System: An idealized dynamic system that ignores disturbances (assuming perfectly accurate wind power forecasting). Nominal MPC solves the optimal control problem on the nominal system, generating a nominal trajectory, which is the centerline of the actual disturbed trajectory.

[0022] Ancillary Feedback Controller: An independent controller in the tube robust MPC (Mechanical Management Circuit), employing linear state feedback (LQR) to compensate for deviations between the actual and nominal systems, ensuring the actual trajectory always remains within the tube centered on the nominal trajectory. In this invention, the auxiliary controller corrects for energy storage power deviations. When the actual SOC deviates from the nominal SOC, it automatically adjusts the charging and discharging power through feedback to pull the SOC back to the tube center.

[0023] Time-Varying Tube Cross-Section Radius: The size of the tube's cross-section at each moment in the prediction time domain determines the safety margin around the nominal trajectory. This invention designs it to adaptively increase with the prediction distance, with a smaller near end (lower conservatism) and a larger far end (higher safety margin), achieving time-varying adaptive robust conservatism.

[0024] Constraint tightening: A technique to pre-tighten nominal system constraints to ensure that the actual disturbed system meets the original hard constraints. The tightening amount is equal to the farthest reach of the pipe cross-section in the constraint direction, ensuring that the actual trajectory meets the original safety constraints at any position within the pipe. In this invention, constraint tightening is applied to all hard safety constraints such as the upper and lower limits of energy storage SOC and the non-negativity of grid-connected power.

[0025] Disturbance Propagation: Wind power prediction errors, as additive disturbances, accumulate continuously through the energy storage SOC integral dynamic channel, resulting in irreversible and cumulative effects on SOC. This is the physical root cause of the expansion of the pipe cross-section along the prediction time domain.

[0026] As described in the background section, the three existing mainstream MPC scheduling schemes cannot simultaneously guarantee strict constraints and achieve scheduling economy, and have the following technical defects: First, fixed deterministic MPC. It is based solely on single-point wind power forecasting optimization and has no safety margin. Rolling corrections have a lag, and sudden wind power fluctuations within the dispatch cycle can easily cause SOC (State of Charge) limits to be exceeded, with the proportion of periods violating constraints reaching 1.6% throughout the year.

[0027] Second, soft constraint penalties for MPC. This indirectly suppresses violations through high fines, representing a statistical-level constraint protection. However, in extreme electricity price scenarios, the optimizer may actively choose to engage in arbitrage beyond limits, lacking mathematically deterministic safety guarantees. The penalty coefficient requires repeated manual adjustments, resulting in poor engineering versatility.

[0028] Third, the classic Minimax Robust MPC. It uniformly tightens constraints across the entire time domain by taking the maximum disturbance interval, ensuring equal conservatism throughout all time periods. However, it fails to match the objective law that wind power error amplifies with the square root of the prediction distance, excessively compressing the energy storage adjustment space in the near-term, potentially resulting in an annual economic loss of over 20%.

[0029] To address the aforementioned issues, this embodiment provides a wind-storage safety scheduling method, namely, proposing a four-dimensional pipe robust model predictive control optimization architecture: "time-varying adaptive pipe cross section—auxiliary linear state feedback—minimum robust positive invariant set offline pre-calculation—pipe boundary adaptive expansion." This architecture achieves deterministic safety assurance across the entire time domain while also considering online low-computing-power deployment capabilities. This method constructs a decoupled dual-component architecture of nominal MPC and auxiliary linear state feedback, designs a time-varying pipe radius with adaptive prediction distance, and a terminal constraint set oriented towards integral dynamics. It provides deterministic constraint guarantees against SOC exceeding limits and power backfeeding under all bounded wind power disturbances, while controlling the economic losses caused by robust conservatism to within 8% of the deterministic MPC, and possesses adaptive fault tolerance capabilities for non-stationary extreme weather.

[0030] For example, such as Figure 2 As shown, this embodiment provides a wind-storage safety scheduling method, including: A discrete linear nominal dynamic model of the wind-storage system is constructed. A bounded additive disturbance set that dynamically changes with the prediction distance is constructed for the wind power prediction deviation. The propagation mapping relationship from wind power disturbance to system state is established, and the system modeling parameters and time-varying disturbance boundary containing the nominal system state equation are obtained. Based on the discrete linear nominal dynamic model of the wind storage system, the deviation dynamic equations of state deviation and control deviation are derived. The feedback gain matrix of the auxiliary linear state feedback controller is solved offline by combining the system modeling parameters. The minimum robust positive invariant set is calculated offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. Combining the reference safety radius parameter of the tube section and the time-varying disturbance boundary, the time-varying tube section radius corresponding to each prediction period is adaptively calculated based on the prediction distance, and the original hard constraints of the system are tightened inward based on the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. Taking the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, the nominal system dynamic constraints are constructed based on the discrete linear nominal dynamic model. The nominal MPC optimization model is built by combining the time-varying tightened constraint set. The optimal nominal control sequence and corresponding nominal state trajectory are obtained by solving the nominal MPC optimization model. The nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period. The feedback gain matrix is ​​invoked, and the real-time state deviation is calculated by combining the real-time collected system state and nominal state trajectory. The current basic scheduling instructions are corrected through linear state feedback, and the actual execution control instructions are generated and issued for execution.

[0031] The wind-storage safety scheduling method provided in this embodiment will be further explained below: like Figure 1 As shown in the figure, this embodiment provides a wind-storage safety scheduling method based on adaptive pipe robust model predictive control. The specific steps are as follows: Step 101: Nominal dynamic modeling and uncertainty disturbance characterization of the wind-storage system: This step establishes the system model and disturbance model upon which the subsequent robust MPC design depends.

[0032] First, construct the nominal dynamic model: using a unified discrete scheduling step size Δt of 30 minutes, construct a linear discrete nominal system for wind power and energy storage (ignoring random errors in wind power prediction). Select the energy storage state of charge and the power at the grid connection point of the power station to form the state vector. x [t]=[SOC[t], P grid [t] T The energy storage charging and discharging power constitutes the control vector. u [t]=[ P ch [t], P dch [t] T Establish a discrete-time linear nominal model: x [t+1]=A x [t]+B u [t]+D d [t], where A is the SOC state transition matrix, B is the charging and discharging control matrix, and D is the exogenous input mapping matrix of wind power and electricity price. All three types of matrices are derived offline from the SOC time-domain recursive equation and the grid-connected power balance equation; exogenous input vector d [t]=[ P wind,pred [t];PriceDA[t] T It consists of two types of deterministic time-series data: NWP wind power forecasts and day-ahead market electricity prices. The system's hard constraints are divided into two categories: the state constraint set X={ x|SOC min ≤SOC≤SOC max , P grid ≥0}, control constraint set U={ u |0≤ P ch ≤ P rated ,0≤ P dch ≤ P rated The maximum charging and discharging power of energy storage in a single time period is limited.

[0033] Second, time-varying bounded disturbance modeling: Actual wind power output has prediction deviations, which satisfy the superposition disturbance relationship: P wind,actual [t+τ]= P wind,pred [t+τ]+w[τ], where the bounded additive perturbation is... w [τ]∈W τ W τ W is a one-dimensional polyhedral perturbation set that dynamically changes with the predicted distance τ. τ ={w||w|≤Δ P (τ)=z α / 2 × s 0×√τ}, where τ is the prediction step size from the current scheduling time to the target time period. s 0 represents the steady-state standard deviation of the single-step wind power prediction error, and the confidence level is... α =0.01, two-tailed standard normal quantile z 0.005 =2.576, the disturbance range can cover 99% of the normal wind power prediction deviation. This formula accurately describes the square root diffusion characteristics of wind power error: for near-end short-term prediction τ=1, the disturbance half-width ΔP(1) is approximately 2.576. s 0; 24-hour forecast at the far end τ=48, perturbation half-width Δ P (48) Approximately 17.81 s 0, the difference in the width of the perturbation interval between the near and far ends is about 7 times. This time-varying perturbation set is the core driving input for predicting the adaptive tube cross-sectional radius and dynamically tightening the constraints in step 103.

[0034] Third, modeling the propagation of disturbance effects on the energy state: Energy storage SOC is an integral dynamic, and wind power disturbances will continuously accumulate in the energy state. The complete recursive expression for discrete SOC is: SOC[t+1] = SOC[t] + (η × ... P ch [t] P dch [t] / or )×Δ t / E ESS Disturbances directly affect the power balance relationship at grid connection points: P grid [t]= P wind,pred [t]+ w [τ]+ P dch [t] P ch [t]. When w [τ]>0, the actual wind power is higher than the forecast. If the energy storage has reached the SOC limit and cannot perform charging operations, and the energy storage discharge power has reached the rated maximum value, the power station cannot fully absorb the excess wind power and can only generate wind curtailment; when w [τ]<0, the actual wind power is lower than the forecast. If the energy storage discharge is insufficient to compensate for the power gap, the SOC is likely to continue to decline and touch the lower limit. The SOC has no internal damping self-stabilizing mechanism. The power deviation caused by a single wind power disturbance will not automatically decay and will continue to accumulate with the forecast time domain. Therefore, the safety margin of the pipe cross-section must be widened synchronously with the forecast distance.

[0035] Step 102: Design of the Auxiliary Linear State Feedback Controller and Offline Calculation of the Minimum Robust Positive Invariant Set: This step performs the calculation of the auxiliary controller gain and tube cross-sectional radius reference offline, and the results can be directly used by the online steps without increasing the online workload.

[0036] First, the dynamic equation of the deviation. Define the state deviation e[t] = x actual [t] x nom [t] and control deviation v[t] = u actual [t] u nom [t], by subtracting the nominal system from the actual system to eliminate the deterministic input terms, the linear deviation dynamic equation is derived: e[t+1]=Ae[t]+Bv[t]+Ew[t], where E is the mapping matrix from wind power disturbance to energy storage state space. Each round of MPC rolling optimization takes the current measured real energy storage state as the nominal trajectory starting point, so the initial deviation of the single round optimization window is e[0]=0.

[0037] Second, the design of the auxiliary feedback controller. The auxiliary correction control adopts the linear state feedback form v[t]=Ke[t]. The feedback gain matrix K is solved offline in one go. The design goal is to make the closed-loop deviation dynamic matrix Acl=A+BK satisfy discrete asymptotic stability, that is, the spectral radius ρ(Acl)<1. The energy storage state matrix A contains the SOC integral channel, and the corresponding eigenvalue is approximately equal to 1 (critical integral mode). Therefore, the spectral radius of the closed-loop matrix can only be compressed to slightly less than 1, and cannot deviate significantly from the unit circle. The spectral radius of A+BK can only be compressed to be close to but slightly less than 1. K is solved by the steady-state solution of the Discrete Algebraic Riccati Equation (DARE). The standard DARE equation is: P=Q+A T PA-A T PB(R+B T PB) 1 B T PA. Solve for the steady-state positive definite solution P using the positive definite weight matrix Q (state penalty, grid connection deviation weight is higher than SOC deviation) and R (control correction penalty), then calculate the LQR feedback gain: K = (R + B) T PB) 1 B T PA.

[0038] Third, offline iterative computation of the minimum robust positive invariant set. mRPI set Ω ∞ Defined as containing the origin and for a fixed worst-case perturbation set W max Satisfying the invariant condition AclΩ ∞ ⊕EW max Ω ∞ The set of minimal convex polyhedra. Ω ∞ The limit achievable range of the state deviation e[t] under all disturbances is characterized, and the support functions of each dimension are the reference safe radius of the pipe section. Due to the wind power disturbance set W... τ As the predicted distance varies over time, to avoid repeatedly solving multiple sets of mRPIs online, a unified set of maximum perturbations W across the entire time domain is selected. max =W τ =48 was used as the offline computation benchmark, and the offline conservatism was moderately increased in exchange for lightweight online computation. The Minkowski summation iterative approximation was adopted to approximate mRPI: Ω0={0}, Ω k+1 =AclΩ k ⊕EW max When the Hausdorff distance is dH(Ω) k+1 ,Ω k )<ε mRPI It converges at time, where ε mRPI To preset the convergence accuracy threshold (1×10 in this embodiment) -5The finite prediction window H=48 limits the maximum number of iterations to 48. In engineering scenarios, convergence is typically achieved in 20-30 iterations. mRPI is calculated offline once before the station is put into operation. After convergence, the polyhedral hyperplane inequality is stored. In online step 103, the support function is directly called to complete the time-varying constraint tightening.

[0039] Step 103: Predicting the adaptive time-varying tube radius and tightening constraints: This step is the core of online calculation, which transforms the offline mRPI benchmark into a pipe radius and constraint tightening amount that are dynamically adjusted according to the predicted distance, achieving an adaptive conservatism of "economical at the near end and safe at the far end".

[0040] First, the calculation of the cross-sectional radius of the time-varying tube. Arbitrary prediction step. t + t The corresponding pipe cross-sectional radius r[ t + t ]equal t Step deviation can reach set Ω τ The maximum value of the support function in the constrained c-normal direction: r c [ t + t ]=max e∈Ωτ h c (e), where Ω τ Let h be the set of state deviations that can be reached after propagating the worst-case perturbation over τ steps from zero initial deviation. c (e) is the support function of the deviation vector e in the direction of constraint c, representing the farthest distance the deviation set extends towards the constraint boundary. For the upper and lower bound constraints of SOC, h SOC (e)=max(e×SOC, e×SOC); for non-negative constraints on grid-connected power, h grid (e)=max( e× P grid The exact calculation of the reachability deviation set involves (0). t The nested Minkowski summations result in computational complexity that increases exponentially with τ. This embodiment proposes a linear interpolation acceleration strategy, utilizing the steady-state mRPI limiting radius r obtained offline in step 102. ∞ End of window t =H corresponds to the support radius r H Constructing the time-varying radius: r c [τ]=min( t / H×r H ,r ∞ This linear interpolation approximates... t When smaller, give a small radius, tAs it approaches H, it tends towards the mRPI radius, which is consistent with the physical process of disturbance accumulation and is extremely simple to calculate.

[0041] Second, the tightening of nominal MPC constraints. The original system's hard constraint set X = {g(x) ≤ 0} needs to be compressed inwards to obtain the nominal optimization tightened feasible region X, ensuring that all disturbed actual states within the system satisfy the original constraints. tight ={g( x nom )≤ r c [τ]×|| g||}, where g is the gradient of the constraint function near the nominal trajectory. In this embodiment, both SOC and grid-connected power are linear constraints, and the gradient is a fixed constant vector; -r c [ t ]×|| g|| represents the safety margin for inward contraction, and the negative sign indicates a shift towards the interior of the feasible region. Specific tightening rules: SOC upper limit constraint SOC nom [ t + t ]≤SOC max r SOC [τ] / E ESS SOC lower bound constraint nom [ t + t ]≥SOC min +r SOC [τ] / E ESS ; Non-negative grid-connected power constraint P grid,nom [ t + t ]≥r grid [ t ]. r SOC [ t [This represents the power deviation margin, divided by the rated energy storage capacity.] E ESS Complete the unit conversion of electricity to dimensionless SOC. The effect of time-varying tightening: For the near end of the forecast, a small τ value allows for a small pipe radius and a low constraint tightening amplitude, resulting in a more relaxed nominal operating range for energy storage, enabling full utilization of capacity for electricity price arbitrage; for the far end... t Approaching H, the pipe radius is significantly widened and the constraint tightening is increased, compressing the nominal range to reserve sufficient safety buffer for severe wind power fluctuations, and realizing an adaptive adjustment mechanism of loose near and tight far.

[0042] Step 104: Nominal MPC Tightening Constraints, Economic Optimization, and Initial Implementation This step solves a compacted small linear programming problem online in each scheduling cycle and outputs the economically optimal nominal control sequence.

[0043] First, construct a nominal MPC optimization model. Each scheduling cycle is based on the current measured energy storage state. x actual [t] serves as the starting point of the rolling time-domain trajectory, forcing x nom [t]= x actual [t], eliminating the initial state deviation of a single round of optimization, construct a finite-time linear programming problem of H=48 (24 hours), with the optimization objective being to minimize the comprehensive operating cost over the entire time period: In the formula J It is a single-period comprehensive cost function, which includes four sub-items: net cost of electricity purchase and sale, marginal depreciation cost of energy storage, wind power forecast assessment penalty, and power deviation recovery cost; J terminal As a terminal penalty term, it guides the time-domain end-point SOC to approach 0.3. E ESS The steady-state objective is to ensure long-term power balance. The model is equipped with four types of constraints: 1) Nominal linear dynamic constraints. x nom [ t + t +1]=A x nom [ t + t ]+B [ t + t ]+D d [ t + t (τ=0,...,H) 1); 2) Time-varying tightening state constraints x nom [ t + t ]∈X tight ( t The constraint set is dynamically generated based on the time-varying tube radius in step 103; 3) Inherent constraints for charge and discharge control. [ t + t ]∈U; 4) Robust terminal invariant set constraint x nom [ t +H]∈X terminal This ensures that the infinite time-domain rolling optimization recursion is feasible and the closed-loop is stable.

[0044] Second, solve the nominal MPC optimization model. This optimization problem is a standard pure linear programming problem. The objective function and all equality / inequality constraints are linear, with no quadratic terms or integer variables. The optimization variables include the charging and discharging power of group 2H, the SOC state of group H, and the power balance auxiliary variables of group H, with a total of approximately 5H = 240 variables and approximately 5H + 1 = 241 constraints. All constraint matrices are stored in CSR compressed sparse row format (sparseness > 98.9%) and directly input into the HiGHS open-source linear solver for calculation, supporting hot-start reuse of basic solutions during the cycle. After the solution is completed, the optimal nominal sequence is output, and the control variables for the first time period are extracted. u nom [t] serves as the current energy storage charge / discharge command; simultaneously, it outputs the complete time-domain nominal state trajectory. x nom [ t +1: t +H] serves as a reference benchmark for the auxiliary feedback controller to calculate the state deviation in step 105.

[0045] Third, regarding the design of the terminal constraint set. Robust terminal invariant set X. terminal Two hard feasibility conditions must be met simultaneously: 1) A stable terminal feedback control law K exists within the set. terminal , any x ∈X terminal After this control is applied, the state in the next moment still belongs to X. terminal 2) The terminal set is completely contained within the furthest tightened feasible region of the prediction window, i.e., X terminal X tight (H) ensures sufficient safety margin for the terminal tubes. In this embodiment, the terminal set is based on a steady-state target SOC = 0.3 × E ESS The elliptic invariant set centered at the center employs an offline joint design process: first, the discrete Lyapunov equations are solved to obtain the fundamental elliptic matrix; then, semidefinite programming is used to maximize the elliptic volume without exceeding the far-end tightening constraints, balancing adjustment space and robust stability. The terminal set is calculated offline once before the station is put into operation. The elliptic constraints obtained from the solution are directly substituted into the nominal MPC of each round as the time-domain end constraint, avoiding the loss of SOC control at the end of the finite time-domain scheduling, and ensuring the feasibility of full-rolling cycle recursion and long-term closed-loop stability.

[0046] Step 105: Online deviation correction and closed-loop state update of auxiliary linear state feedback: First, online calculation of auxiliary control. After each round of nominal MPC solution, the actual energy storage control quantity is obtained by combining nominal optimal control and linear feedback correction: u actual [t]=u nom [t]+K×( x actual [t] x nom [t]), where the feedback gain matrix K is the LQR gain obtained by offline DARE solution in step 102; x nom [t] represents the initial nominal state output by this round of MPC. Since each round of optimization forces the real-time measured energy storage state as the trajectory starting point, it satisfies... x actual [ t ]= x nom [ t When the instantaneous state deviation e[t] = 0, the initial correction amount is 0, and the nominal control is directly used as the initial output. The core value of the auxiliary linear feedback is reflected in the 30-minute scheduling window between the two MPC solutions: during the window, wind power real-time disturbances continue to act, and without continuous feedback correction, the SOC deviation will continue to accumulate along the integral channel; the auxiliary controller corrects the charging and discharging power in real time at high frequency, continuously compressing the deviation expansion rate and delaying the state from touching the pipe boundary. After the SCADA collects complete operating condition data in the next scheduling cycle, the MPC resets the nominal trajectory starting point with the latest real SOC, clears all deviations caused by the disturbances in this round to zero, and restarts the safety pipeline from zero.

[0047] Second, disturbance monitoring and adaptive scaling for boundary violations. At the end of each 30-minute complete scheduling cycle, SCADA uploads the actual wind power output for that period. P wind,actual [t] is compared with the measured SOC at the terminal, and the state deviation at the end of this round is calculated as: e[t+1] = x actual [t+1] x nom [t+1]. If the absolute value of the SOC deviation is |e SOC [t+1]|>r SOC [t+1] or negative deviation of grid-connected power |e Pgrid [t+1]|>r grid [t+1] indicates that the trigger tube boundary violation warning has been issued, meaning that the actual wind power disturbance exceeds the reference disturbance set W corresponding to the offline mRPI. max,0 An adaptive adjustment strategy is adopted: the upper limit W of the global perturbation set is increased. maxTemporarily increase by 20%, with the upper limit after increase not exceeding 1.5 times the original baseline. In the next scheduling cycle, step 103, the radius of all time-varying tubes is simultaneously widened, and the constraint tightening amount is simultaneously increased. If there are no boundary warnings for three consecutive complete scheduling cycles, W will be increased in each cycle. max Reduce by 5%, and the minimum must not be lower than the original offline baseline W. max,0 The system gradually restores its normal conservatism. This adaptive mechanism enables the pipe radius to automatically expand during extreme weather events (disturbances exceeding historical statistical ranges) and automatically shrink after the extreme events end, achieving pipe MPC adaptation to non-stationary disturbance environments with extremely low computational cost (adjusting only one scalar parameter).

[0048] Step 106: Full-cycle rolling iteration and year-round counterfactual performance evaluation: First, the full-cycle rolling iterative process. After a single round of scheduling is completed, the time sequence index t = t + 1 is set, and the actual energy storage state is obtained by SCADA measurement in step 105. x actual [t+1] serves as the initial starting point for the nominal trajectory of the next round of MPC. The complete closed-loop sequence is as follows: Step 103 (Read the adaptively updated W) max (Calculate the time-varying tube radius and hierarchical tightening constraints) → Step 104 (Solve the nominal linear programming with 240 variables and output the optimal nominal control sequence) → Issue the first charging and discharging command to execute a 30-minute scheduling cycle → After the cycle ends, proceed to step 105 to collect actual wind power and SOC, calculate the state deviation, and adaptively scale the disturbance set W as needed. max After completing one closed loop, return to step 103 for iteration.

[0049] Second, offline counterfactual quantification performance evaluation. Based on a complete year-round measured dataset of wind power output, electricity price, and energy storage operation sequence at the power station, a unified simulation environment was built. Counterfactual simulations were conducted for four types of comparative strategies throughout the year and all time periods. The four benchmark schemes were: fixed deterministic MPC, penalty-constrained MPC, classic all-time conservative robust MPC, and the adaptive pipe robust MPC provided in this embodiment. Simulation results show that the safety performance of this adaptive pipe robust MPC is completely comparable to that of the classic conservative robust MPC, with near-zero violations at full SOC and grid-connected backfeed. Compared to the classic robust MPC, the annual arbitrage return is increased by more than 14%, while the economic loss caused by robustness is controlled within 8%. It achieves balanced optimization at the safety-return Pareto optimal frontier, verifying the comprehensive advantages of the entire six-step coupled adaptive architecture.

[0050] For example, the wind-storage safety scheduling method provided in this embodiment is applied and implemented, and the specific implementation process is as follows: Construct a robust MPC architecture with nominal MPC economic optimization and LQR-assisted feedback dual decoupling; the half-width of the wind power disturbance set increases with the square root of the prediction distance; adopt...r c [ t ]=min( t / H× r H , r ∞ Linear interpolation adaptive pipe radius dynamic tightening constraint; real-time monitoring of deviation between actual state and nominal trajectory, when the pipe boundary crosses the boundary, the disturbance set automatically expands by 20%, and shrinks by 5% cycle by cycle after 3 consecutive cycles without abnormality; offline coupled solution of LQR gain and mRPI minimum robust invariant set, design of robust terminal constraint set based on Lyapunov equation, achieving SOC and grid-connected power deterministic zero violation under 99% normal disturbance, and annual economic loss is less than 8%.

[0051] Specifically, wind power time-varying disturbance set modeling: wind power additive disturbances satisfy W τ ={ w| |w|≤Δ P ( t )=z α / 2 × s 0×√τ},z 0.005 =2.576 covers 99% of the prediction error, where τ is the rolling time-domain prediction distance. s 0 represents the standard deviation of the one-step wind power prediction error. The disturbance half-width increases synchronously with the prediction distance, matching the characteristic of increased long-term power output fluctuations.

[0052] Furthermore, the offline design process for LQR coupled with mRPI is as follows: Construct the deviation dynamic equation, solve the DARE equation to obtain the linear feedback gain K, and set the grid-connected power deviation penalty weight higher than the SOC deviation; verify that the spectral radius of the closed-loop matrix is ​​less than 1; calculate the minimum robust positive invariant set mRPI through Minkowski summation iteration, and set the convergence accuracy threshold ε. mRPI Take 1×10 -5 Offline storage tube reference radius r ∞ Window end radius r H Available for online use.

[0053] Among them, adopt r c [ t ]=min( t / H× r H , r ∞ Linear interpolation is used to calculate the radius of each time-domain tube section; based on the radius, the upper and lower limits of SOC are tightened synchronously, and the original hard constraint of grid connection is non-negative. The tightening amplitude is small at the near end and large at the far end, and the approximate error of linear interpolation is controlled within 8%.

[0054] In this embodiment, the pipe boundary over-boundary adaptive disturbance set adjustment mechanism is set as follows: the deviation norm between the actual SOC and the nominal trajectory is calculated every scheduling cycle; if the deviation exceeds the current pipe radius, the upper limit of the global disturbance set is expanded by 20%, and the constraint margin of the entire time domain is widened simultaneously; when there are no boundary warnings for 3 consecutive cycles, the disturbance set is reduced by 5% every cycle to gradually restore the baseline safe pipeline.

[0055] Furthermore, the robust terminal constraint set is specifically designed offline as follows: the fundamental matrix of the elliptic invariant set is solved based on the Lyapunov equation, and combined with semidefinite programming to expand to the maximum volume that the tight constraint can accommodate. The terminal set ensures that the rolling MPC has a feasible solution throughout the entire process, realizing stable closed-loop scheduling in the infinite time domain.

[0056] As another preferred embodiment, the offline counterfactual comparison and evaluation framework is designed as follows: using complete scheduling data for the whole year, four types of strategies are compared: deterministic MPC, penalized MPC, classic conservative robust MPC, and local adaptive robust MPC. The total number of constraint violations, the total annual energy storage arbitrage revenue, the average solution time per cycle, and the scheduling performance ratio PR are used as four-dimensional quantitative evaluation indicators.

[0057] In summary, this embodiment provides a wind-storage safety scheduling method with the following advantages: First, the application of robust model predictive control in wind and energy storage scheduling and its dual-component decoupled architecture. A dual-component architecture of "nominal MPC (tightening constraint LP economic optimization) + auxiliary LQR feedback (disturbance suppression and safety assurance)" is constructed to achieve separation of responsibilities and independent optimization of economic and safety, and to ensure SOC and grid connection hard constraints in terms of mathematical rigor.

[0058] Second, the time-varying tube cross-sectional radius and constraint tightening are adaptive to the predicted distance. The tube radius increases linearly from the near end to the far end mRPI value as the predicted distance increases, and the constraint tightening amount changes synchronously with the time, achieving adaptive robustness of "low conservatism at the near end to make full use of energy storage and high conservatism at the far end to reserve disturbance buffer".

[0059] Third, an integral SOC dynamic terminal constraint set design is adopted. Considering the characteristic of disturbances continuously accumulating in the SOC integral channel, an elliptic terminal constraint set centered on the target SOC is designed. This set is solved offline using Lyapunov equations and LMI semidefinite programming to ensure the recursive feasibility of nominal MPC and the ability to continue operating in the infinite time domain.

[0060] Fourth, linear interpolation for the τ-step reachable deviation set is accelerated. Utilizing the physical property that bounded perturbations accumulate linearly in the integration channel, linear interpolation replaces exact Minkowski summation, reducing the complexity of online tube radius calculation.

[0061] Fifth, boundary violation detection and adaptive expansion / contraction of the disturbance set. The system monitors the deviation between the actual state and the nominal trajectory online. When the deviation exceeds the boundary, the disturbance set is automatically expanded. After three consecutive cycles without violation, the set is gradually contracted, achieving robust fault tolerance for non-stationary extreme weather conditions through a single scalar parameter adaptive mechanism.

[0062] Sixth, it adopts a two-layer lightweight solution architecture of "constrained tightening LP + LQR feedback". The online computation is decomposed into a 240-variable sparse LP and a single matrix multiplication, with a computation cycle in the order of seconds, far less than the 30-minute scheduling interval, and can be reliably deployed on edge computing hardware.

[0063] For example, such as Figure 3 As shown, this embodiment also provides a wind-storage safety dispatching system, including: The model building module is used to construct a discrete linear nominal dynamic model of the wind-storage system, construct a bounded additive disturbance set that dynamically changes with the prediction distance for wind power prediction deviation, and establish the propagation mapping relationship from wind power disturbance to system state, thereby obtaining system modeling parameters and time-varying disturbance boundaries that include the nominal system state equation. The derivation and solution module is used to derive the deviation dynamic equations of state deviation and control deviation based on the discrete linear nominal dynamic model of the wind storage system, solve the feedback gain matrix of the auxiliary linear state feedback controller offline in combination with the system modeling parameters, and calculate the minimum robust positive invariant set offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. The calculation module is used to combine the reference safety radius parameter of the tube section and the time-varying disturbance boundary, adaptively calculate the time-varying tube section radius corresponding to each prediction period according to the prediction distance, and perform inward tightening processing on the original hard constraints of the system according to the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. The instruction extraction module is used to take the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, construct the nominal system dynamic constraints based on the discrete linear nominal dynamic model, build the nominal MPC optimization model in combination with the time-varying tightened constraint set, solve the nominal MPC optimization model to obtain the optimal nominal control sequence and the corresponding nominal state trajectory, and extract the nominal control quantity of the first time period as the basic scheduling instruction for the current period. The feedback correction module is used to call the feedback gain matrix, calculate the real-time state deviation by combining the real-time collected system state and nominal state trajectory, correct the current basic scheduling instructions through linear state feedback, generate actual execution control instructions and issue them for execution.

[0064] The present invention also provides a wind and energy storage safety scheduling device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the wind and energy storage safety scheduling method.

[0065] The present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the wind and energy storage safety scheduling method.

[0066] When the processor executes the computer program, it implements the above-mentioned steps for wind and energy storage safety scheduling, such as: constructing a discrete linear nominal dynamic model of the wind and energy storage system, constructing a bounded additive disturbance set that dynamically changes with the prediction distance for wind power prediction deviation, and establishing a propagation mapping relationship from wind power disturbance to system state, thereby obtaining system modeling parameters and time-varying disturbance boundaries that include the nominal system state equation. Based on the discrete linear nominal dynamic model of the wind storage system, the deviation dynamic equations of state deviation and control deviation are derived. The feedback gain matrix of the auxiliary linear state feedback controller is solved offline by combining the system modeling parameters. The minimum robust positive invariant set is calculated offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. Combining the reference safety radius parameter of the tube section and the time-varying disturbance boundary, the time-varying tube section radius corresponding to each prediction period is adaptively calculated based on the prediction distance, and the original hard constraints of the system are tightened inward based on the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. Taking the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, the nominal system dynamic constraints are constructed based on the discrete linear nominal dynamic model. The nominal MPC optimization model is built by combining the time-varying tightened constraint set. The optimal nominal control sequence and corresponding nominal state trajectory are obtained by solving the nominal MPC optimization model. The nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period. The feedback gain matrix is ​​invoked, and the real-time state deviation is calculated by combining the real-time collected system state and nominal state trajectory. The current basic scheduling instructions are corrected through linear state feedback, and the actual execution control instructions are generated and issued for execution.

[0067] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing preset functions, wherein the instruction segments describe the execution process of the computer program in the wind-storage safety scheduling device. For example, the computer program can be divided into a model building module, a derivation and solution module, a calculation module, an instruction extraction module, and a feedback correction module; the specific functions are as follows: The model building module is used to build a discrete linear nominal dynamic model of the wind-storage system, construct a bounded additive disturbance set that dynamically changes with the prediction distance for wind power prediction deviation, and establish the propagation mapping relationship from wind power disturbance to system state, obtaining system modeling parameters and time-varying disturbance boundaries containing the nominal system state equation; The derivation and solution module is used to derive the deviation dynamic equations of state deviation and control deviation based on the discrete linear nominal dynamic model of the wind-storage system, solve the feedback gain matrix of the auxiliary linear state feedback controller offline in combination with the system modeling parameters, and calculate the minimum robust positive invariant set offline for the maximum disturbance set in the full time domain, obtaining the reference safety radius parameter of the pipe section; The calculation module is used to combine the reference safety radius parameter of the pipe section The system employs a time-varying perturbation boundary calculation method. Based on the predicted distance, it adaptively calculates the radius of the time-varying tube cross-section corresponding to each prediction period and performs inward tightening processing on the original hard constraints of the system according to the time-varying tube cross-section, generating a time-varying tightened constraint set corresponding to each prediction period. The instruction extraction module uses the current measured actual operating state of the wind-storage system as the initial point for rolling optimization. It constructs nominal system dynamic constraints based on a discrete linear nominal dynamic model, builds a nominal MPC optimization model in combination with the time-varying tightened constraint set, solves the nominal MPC optimization model to obtain the optimal nominal control sequence and corresponding nominal state trajectory, and extracts the nominal control quantity of the first period as the current basic scheduling instruction. The feedback correction module calls the feedback gain matrix, calculates the real-time state deviation in combination with the real-time collected system state and nominal state trajectory, corrects the current basic scheduling instruction through linear state feedback, generates actual execution control instructions, and issues them for execution.

[0068] The wind and energy storage safety scheduling device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The wind and energy storage safety scheduling device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the above are examples of wind and energy storage safety scheduling devices and do not constitute a limitation on them. The device may include more components than described above, or combine certain components, or use different components. For example, the wind and energy storage safety scheduling device may also include input / output devices, network access devices, buses, etc.

[0069] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or any conventional processor, etc. The processor is the control center of the wind and energy storage safety dispatching system, connecting various parts of the entire wind and energy storage safety dispatching equipment via various interfaces and lines.

[0070] The memory can be used to store the computer program and / or modules. The processor implements various functions of the wind and energy storage safety scheduling device by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory.

[0071] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital cards (SD cards), flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0072] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the wind and energy storage safety scheduling method.

[0073] If the modules / units integrated in the wind and energy storage safety dispatch system are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0074] Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned wind and energy storage safety scheduling method, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-mentioned wind and energy storage safety scheduling method. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0075] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0076] It should be noted that the content contained in the computer-readable storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0077] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A wind-storage safety dispatching method, characterized in that, include: A discrete linear nominal dynamic model of the wind-storage system is constructed. A bounded additive disturbance set that dynamically changes with the prediction distance is constructed for the wind power prediction deviation. The propagation mapping relationship from wind power disturbance to system state is established, and the system modeling parameters and time-varying disturbance boundary containing the nominal system state equation are obtained. Based on the discrete linear nominal dynamic model of the wind storage system, the deviation dynamic equations of state deviation and control deviation are derived. The feedback gain matrix of the auxiliary linear state feedback controller is solved offline by combining the system modeling parameters. The minimum robust positive invariant set is calculated offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. Combining the reference safety radius parameter of the tube section and the time-varying disturbance boundary, the time-varying tube section radius corresponding to each prediction period is adaptively calculated based on the prediction distance, and the original hard constraints of the system are tightened inward based on the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. Taking the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, the nominal system dynamic constraints are constructed based on the discrete linear nominal dynamic model. The nominal MPC optimization model is built by combining the time-varying tightened constraint set. The optimal nominal control sequence and corresponding nominal state trajectory are obtained by solving the nominal MPC optimization model. The nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period. The feedback gain matrix is ​​invoked, and the real-time state deviation is calculated by combining the real-time collected system state and nominal state trajectory. The current basic scheduling instructions are corrected through linear state feedback, and the actual execution control instructions are generated and issued for execution.

2. The wind-storage safety dispatching method according to claim 1, characterized in that, The construction of the discrete linear nominal dynamic model of the wind-storage system includes: The system state vector is formed by selecting the energy storage state of charge and the power at the grid connection point of the power station; the control vector is formed by selecting the energy storage charging power and discharging power; and the exogenous input vector is formed by selecting the wind power forecast power and the day-ahead market electricity price. The discrete-time linear nominal system state equation is established to obtain the discrete linear nominal dynamic model of the wind-storage system. The bounded additive perturbation set is constructed based on the normal distribution statistical characteristics of wind power prediction errors, and the width of the perturbation interval increases with the square root of the prediction step. The propagation mapping relationship is used to characterize the integral cumulative effect of wind power disturbance in the state of charge of energy storage, as well as the direct superposition effect of wind power disturbance on the power at the grid connection point.

3. The wind-storage safety dispatching method according to claim 1, characterized in that, The process of solving the reference safety radius parameter of the pipe cross section includes: By subtracting the actual system from the nominal system to eliminate deterministic input terms, the linear deviation dynamic equation is obtained. The feedback gain matrix is ​​solved; the optimal linear quadratic feedback gain matrix is ​​obtained by solving the steady-state solution of the discrete algebraic Riccati equation, so that the closed-loop deviation dynamic system satisfies discrete asymptotic stability. The maximum perturbation set in the entire prediction time domain is selected as the calculation benchmark. The minimum robust positive invariant set is calculated by the Minkowski summation iterative approximation method. The support function values ​​of each dimension of the convex polyhedron set obtained after iterative convergence are used as the benchmark safety radius parameter of the pipe section.

4. The wind-storage safety dispatching method according to claim 1, characterized in that, The adaptive calculation of the time-varying tube cross-sectional radius corresponding to each prediction period based on the predicted distance includes: Linear interpolation is used for calculation, so that the near-end prediction period corresponds to a smaller pipe cross-section radius, and the far-end prediction period corresponds to a larger pipe cross-section radius that is close to the pipe cross-section reference safety radius; The original hard constraints of the system are tightened inward. Specifically, the constraint tightening amount is calculated based on the time-varying tube cross-sectional radius and the constraint function gradient. The original state constraints are shrunk into the feasible region to obtain the time-varying tightened constraint set corresponding to each prediction period, forming an adaptive constraint adjustment mechanism that is loose in the near term and tight in the far term.

5. The wind-storage safety dispatching method according to claim 1, characterized in that, The nominal MPC optimization model aims to minimize the comprehensive operating cost within the finite scheduling time domain. The comprehensive operating cost includes the net cost of electricity purchase and sale, the cost of energy storage operation attenuation, and the cost of power deviation assessment. The nominal MPC optimization model is also equipped with inherent constraints on charging and discharging power and robust terminal invariant set constraints. After solving for the optimal nominal control sequence, the nominal control quantity of the first time period is extracted as the basic scheduling instruction for the current period.

6. The wind-storage safety dispatching method according to claim 1, characterized in that, Each round of nominal MPC optimization forces the current measured actual operating state of the wind-storage system as the starting point of the nominal trajectory, eliminating the initial state deviation of a single round of optimization. Within a single scheduling cycle, the energy storage charging and discharging power is continuously corrected through linear state feedback to suppress the continuous expansion of the state deviation caused by wind power disturbances along the energy storage charge state integral channel. After the end of each scheduling cycle, the starting point of the next round of nominal trajectory is reset with the latest measured actual system state, and the accumulated state deviation of the previous cycle is cleared.

7. The wind-storage safety dispatching method according to claim 1, characterized in that, At the end of each scheduling cycle, the actual wind power output and energy storage state of charge data are collected to calculate the state deviation at the end of the cycle. When the state deviation at the end of the cycle is detected to exceed the cross-sectional radius of the time-varying tube in the corresponding time period, it is determined that the tube boundary violation warning is triggered, the upper limit of the global disturbance set is amplified according to the preset ratio, and the cross-sectional radius of the time-varying tube and the time-varying tightening constraint set of each time period are updated simultaneously. When no boundary violation warning occurs for several consecutive scheduling cycles, the upper limit of the disturbance set is gradually reduced to the baseline value calculated offline according to a preset ratio, and the scheduling process is executed cyclically to achieve full-cycle rolling optimization.

8. A wind-storage safety dispatching system, characterized in that, include: The model building module is used to construct a discrete linear nominal dynamic model of the wind-storage system, construct a bounded additive disturbance set that dynamically changes with the prediction distance for wind power prediction deviation, and establish the propagation mapping relationship from wind power disturbance to system state, thereby obtaining system modeling parameters and time-varying disturbance boundaries that include the nominal system state equation. The derivation and solution module is used to derive the deviation dynamic equations of state deviation and control deviation based on the discrete linear nominal dynamic model of the wind storage system, solve the feedback gain matrix of the auxiliary linear state feedback controller offline in combination with the system modeling parameters, and calculate the minimum robust positive invariant set offline for the maximum disturbance set in the whole time domain to obtain the reference safe radius parameter of the pipe section. The calculation module is used to combine the reference safety radius parameter of the tube section and the time-varying disturbance boundary, adaptively calculate the time-varying tube section radius corresponding to each prediction period according to the prediction distance, and perform inward tightening processing on the original hard constraints of the system according to the time-varying tube section to generate the time-varying tightening constraint set corresponding to each prediction period. The instruction extraction module is used to take the current measured actual operating state of the wind-storage system as the initial point for rolling optimization, construct the nominal system dynamic constraints based on the discrete linear nominal dynamic model, build the nominal MPC optimization model in combination with the time-varying tightened constraint set, solve the nominal MPC optimization model to obtain the optimal nominal control sequence and the corresponding nominal state trajectory, and extract the nominal control quantity of the first time period as the basic scheduling instruction for the current period. The feedback correction module is used to call the feedback gain matrix, calculate the real-time state deviation by combining the real-time collected system state and nominal state trajectory, correct the current basic scheduling instructions through linear state feedback, generate actual execution control instructions and issue them for execution.

9. A wind-storage safety dispatching device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the wind-storage safety scheduling method according to any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the wind and energy storage safety scheduling method according to any one of claims 1-7.

Images