A sequential linearization method for optimal control of grid-forming energy storage

Abstract

The application provides a sequential linearization solving method and device for network-constructed energy storage optimal control problems, comprising: establishing a system frequency dynamic characteristic representation model and an energy storage dynamic characteristic representation model; constructing a model prediction optimal control framework for network-constructed energy storage, considering the coupling characteristics of system frequency dynamics and energy storage dynamics, and optimizing the optimal control strategy under the minimum cost of network-constructed energy storage; analyzing the linearization constraint conditions of the model prediction optimal control framework, and realizing the linearization representation of the system frequency dynamics and the energy storage dynamics; analyzing the linearization objective function of the model prediction optimal control framework, and realizing the linearization mapping between the optimization objective and the system state and the energy storage state; constructing a sequential linearization solving framework, converting the nonlinear optimization problem of the optimal control framework into a group of iterative and progressive linear optimization problems, and realizing the fast and stable convergence of the algorithm through an acceptance rejection mechanism and a credible domain updating mechanism.

Description

Technical Field

[0001] This invention belongs to the field of optimal control technology for grid-type energy storage, and specifically relates to a sequential linearization solution method for the optimal control problem of grid-type energy storage. Background Technology

[0002] Currently, grid-based energy storage is a key stable support resource in the context of large-scale integration of new energy sources into the new power system. Grid-based energy storage requires reasonable control methods to adjust its grid-connection capacity, namely, inertia and damping settings. How to achieve optimal control of grid-based energy storage while minimizing its cost and maintaining system frequency stability is one of the current main focuses.

[0003] Optimal control of grid-connected energy storage involves dynamically adjusting internal inertia and damping settings with the goal of ensuring system frequency stability as the boundary condition. Due to the nonlinear characteristics of the system frequency response, this type of problem is typically a nonlinear optimization problem with high computational complexity. Researchers from ETH Zurich and the University of Leeds have constructed a virtual synchronous machine control algorithm based on a linear quadratic regulator to dynamically and optimally adjust the state feedback loop of inertia and damping under system frequency disturbances. Researchers from Northeastern University, Tsinghua University Shenzhen International Research Institute, and Aalborg University have established a mapping relationship between system deviation and the grid-connected energy storage capacity. Based on offline data sampling analysis, they utilize adaptive dynamic programming and echo state networks to learn this mapping relationship in a data-driven manner and apply it to real-time online grid-connected energy storage control. Researchers from the University of Washington have constructed a model predictive control algorithm oriented towards system frequency constraints, explicitly considering the grid frequency stability requirements. Based on a rolling optimization framework, they correct the system frequency deviation under real-time disturbances for the system frequency response model to obtain the control quantity of the grid-connected devices under optimal control.

[0004] The main problems facing optimal control of existing grid-based energy storage systems stem from two aspects: First, insufficient consideration of the internal dynamic characteristics of energy storage makes it impossible to measure the cost of grid-based energy storage and guarantee the feasibility of control strategies in physical entities. Second, the dynamic consideration of system frequency is overly simplified, with most existing methods assuming ideal step disturbances. However, in actual operation, real-time disturbances are continuously changing, and the simplified paths adopted by existing methods often overestimate the grid-based capacity required by the energy storage system, resulting in unnecessary economic losses and additional stability problems. To address these shortcomings, a complete optimal control model should be proposed, but this model is highly nonlinear and difficult to solve.

[0005] Therefore, from the perspective of existing technical solutions, there is still a lack of optimal control solution methods and devices that are adapted to real-time disturbances, taking into account the dynamic coupling characteristics of system frequency and energy storage. Summary of the Invention

[0006] The present invention aims to at least partially solve one of the technical problems in the related art.

[0007] Therefore, the first objective of this invention is to propose a sequential linearization solution method for the optimal control problem of grid-type energy storage.

[0008] The second objective of this invention is to propose a sequential linearization solution device for the optimal control problem of grid-type energy storage.

[0009] The third objective of this invention is to provide a computer device.

[0010] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0011] To achieve the above objectives, a first aspect of the present invention proposes a sequential linearization solution method for the optimal control problem of grid-type energy storage, comprising:

[0012] A system frequency dynamic characteristic characterization model is established to characterize the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. Establish a dynamic characteristic characterization model for energy storage to depict the state changes and output changes of grid-type energy storage during dynamic operation; Based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model, a model prediction optimal control framework for grid-type energy storage is constructed. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The analytical model predicts the linearized constraints of the optimal control framework, realizing the linearized characterization of system frequency dynamics and energy storage dynamics; The linearized objective function of the analytical model predicting the optimal control framework is linearized to achieve a linear mapping between the optimization objective and the system state and energy storage state; A sequential linearization solution framework is constructed, which transforms the nonlinear optimization problem of the model predicting the optimal control framework into a set of iterative linear optimization problems for solution. The algorithm achieves fast and stable convergence through acceptance / rejection mechanisms and a credibility region update mechanism.

[0013] In one embodiment of the present invention, the constructed system frequency dynamic characteristic characterization model is based on discrete state space, adopts a model predictive control framework, defines the sampling step size, control step size and prediction interval, and discretizes the continuous state space model by the backward Euler method. The system frequency dynamic characteristic characterization model has the following form:

[0014] in This represents the derivative of the system state vector with respect to time. and The system state matrix is ​​determined by the control variables of the grid-type energy storage. This represents the control parameters for grid-type energy storage, including the virtual inertia and virtual damping of grid-type devices. This represents the load fluctuation vector at each node. The system state vector includes the grid phase angle offset, grid angular frequency offset, and synchronous generator mechanical power offset at each node in the grid node set.

[0015] In one embodiment of the present invention, the energy storage dynamic characteristic characterization model includes the state update relationship between the energy storage state of charge and the polarization voltage, and considers the nonlinear mapping relationship between the open circuit voltage of the energy storage cell and the state of charge, as well as the impact of inverter efficiency loss on output power. The energy storage dynamic characteristic characterization model has the following form:

[0016] in This represents the derivative of the energy storage state vector with respect to time. Indicates the energy storage input current. and The time-invariant state transition matrix is ​​determined by the parameters of the grid-type energy storage cells. This represents the energy storage state vector, which includes the state of charge and polarization voltage.

[0017] In one embodiment of the present invention, the model predictive optimal control framework incorporates the frequency stability requirement into the objective function in the form of a penalty function, and achieves feasible control commands under large disturbances by penalizing the frequency offset exceeding the limit. The optimization problem of the model prediction optimal control framework for the grid-type energy storage is as follows:

[0018] In the formula Describe the objective function. To exceed the limit penalty coefficient, These are the upper and lower limits of the permissible state of charge (SOC) for grid-type energy storage operation, respectively. These are the upper and lower limits of the permissible current during the operation of grid-type energy storage. This represents the rated feasible range for grid-type energy storage control.

[0019] In one embodiment of the present invention, the linearization process of the constraint conditions includes: The optimization variables are reconstructed by introducing auxiliary control variables; Taylor expansion is performed on the system frequency dynamic characteristic characterization model near the original optimization solution to obtain the linearized system state and output relationship; The open-circuit voltage curve of the energy storage cell is approximated by a piecewise linear approach; the equivalent current on the grid side is introduced as an auxiliary variable to linearize the coupling relationship between the output power of the grid-type energy storage and the power of the internal cells.

[0020] In one embodiment of the present invention, the linearization process of the objective function includes: Introduce a non-negative auxiliary variable to equivalently replace the truncation treatment that exceeds the frequency limit; By introducing an auxiliary variable to replace the calculation of the absolute value of power, the energy storage cost can be expressed in a linear form.

[0021] In one embodiment of the present invention, the iterative process of the sequence linearization solution framework includes: The original problem is linearized near the current optimal solution to obtain a linear optimization subproblem; this linear optimization subproblem is solved, and the acceptance / rejection mechanism is used to determine whether to accept the new solution; the confidence region size is dynamically updated based on the acceptance or rejection result; if the linear optimization subproblem has no feasible solution, the linearized relaxation problem with relaxation variables is solved to obtain the search direction of the regression feasible region; The acceptance / rejection mechanism determines whether to accept a new solution by comparing the decrease value of the linearized objective function with the decrease value of the actual objective function and combining it with a preset minimum acceptance rate; the trust region update mechanism expands or shrinks the trust region range proportionally based on the acceptance status of the current solution.

[0022] To achieve the above objectives, a second aspect of the present invention provides a sequential linearization solution device for the optimal control problem of grid-type energy storage, comprising: The system frequency modeling module is used to establish a dynamic characteristic model of the system frequency, and to characterize the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. The energy storage dynamic modeling module is used to establish a dynamic characteristic model of energy storage and to characterize the state changes and output changes of grid-type energy storage during dynamic operation. The optimization framework construction module is used to construct the model prediction optimal control framework for grid-type energy storage based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The linearization module is used to analyze the linearized constraints of the model predicting the optimal control framework, and realize the linearized representation of the system frequency dynamics and energy storage dynamics; it analyzes the linearized objective function of the model predicting the optimal control framework and performs linearization processing to realize the linear mapping between the optimization objective and the system state and energy storage state. The sequence solving module is used to execute the sequence linearization solution framework. It solves the original nonlinear optimization problem by iteratively solving the linear optimization subproblems, and supports acceptance / rejection mechanisms and credibility region update mechanisms.

[0023] This invention presents a sequential linearization solution method and apparatus for the optimal control problem of grid-type energy storage. Compared with previous methods, this invention innovatively proposes a sequential linearization solution method and apparatus for the optimal control problem of grid-type energy storage based on the coupling characteristics of system frequency dynamics and energy storage dynamics, and considering real-time disturbances. First, this invention accurately models the system frequency dynamics and energy storage dynamics, and connects the coupling characteristics of grid-type energy storage operation under real-time disturbances using a power interface. Second, this invention proposes a linearization method for optimal control problems with highly nonlinear coupling characteristics, linearizing and approximating the constraints and objective function of the original problem. Finally, this invention designs a sequential linearization solution framework, rapidly iterating to obtain the optimal solution of the original problem through an iterative loop of linearization → solution → update. The proposed scheme significantly improves the solution efficiency of optimal control of grid-type devices with highly nonlinear characteristics. Furthermore, the mathematical foundation of the proposed solution method is the relatively mature linear optimization, which can obtain stable and reliable optimal solutions, improving the optimality of the final solution. This invention can improve the frequency stability capability of grid-based energy storage support systems, reduce the operating costs of grid-based energy storage, avoid redundant resource consumption, and promote the development of high-proportion renewable energy power systems. Therefore, this invention has significant practical implications and broad application prospects.

[0024] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing a sequential linearization solution method for a grid-type energy storage optimal control problem as described in the first aspect embodiment.

[0025] To achieve the above objectives, a fourth aspect of this application proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a sequential linearization solution method for the optimal control problem of grid-type energy storage as described in the first aspect embodiment.

[0026] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0027] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a sequential linearization solution method for the optimal control problem of network-type energy storage according to an embodiment of the present invention; Figure 2 This is a structural diagram of a sequential linearization solution device for the optimal control problem of network-type energy storage according to an embodiment of the present invention; Figure 3 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] The following describes, with reference to the accompanying drawings, a sequential linearization solution method and apparatus for the optimal control problem of grid-type energy storage according to an embodiment of the present invention.

[0031] Figure 1 This is a flowchart of a sequential linearization solution method for the optimal control problem of network-type energy storage according to an embodiment of the present invention, as shown below. Figure 1 As shown, it includes: A system frequency dynamic characteristic characterization model is established to characterize the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. Establish a dynamic characteristic characterization model for energy storage to depict the state changes and output changes of grid-type energy storage during dynamic operation; Based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model, a model prediction optimal control framework for grid-type energy storage is constructed. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The analytical model predicts the linearized constraints of the optimal control framework, realizing the linearized characterization of system frequency dynamics and energy storage dynamics; The linearized objective function of the analytical model predicting the optimal control framework is linearized to achieve a linear mapping between the optimization objective and the system state and energy storage state; A sequential linearization solution framework is constructed, which transforms the nonlinear optimization problem of the model predicting the optimal control framework into a set of iterative linear optimization problems for solution. The algorithm achieves fast and stable convergence through acceptance / rejection mechanisms and a credibility region update mechanism.

[0032] The present invention provides a sequential linearization solution method for the optimal control problem of grid-type energy storage. This method transforms the original problem into a set of linearized problems that can be solved efficiently. Based on the coupling characteristics of system frequency dynamics and energy storage dynamics, it improves the accuracy and response speed of grid-type energy storage control in response to real-time disturbances.

[0033] The following section, with reference to the accompanying drawings, details a sequential linearization solution method for the optimal control problem of grid-type energy storage according to an embodiment of the present invention.

[0034] The sequential linearization solution method and apparatus for the optimal control problem of grid-type energy storage of the present invention are specifically based on model predictive control implemented in discrete state space, defining the sampling step size as... The sampling point labels are Control step size is ( ), control point labels are The prediction interval is ( ).

[0035] The method includes the following steps: 1) Establish a system frequency dynamic characteristic characterization model to depict the system frequency response under real-time disturbances; 2) Establish a dynamic characteristic characterization model for energy storage to depict the state changes and output changes during the dynamic operation of grid-type energy storage; 3) Construct a model prediction optimal control framework for grid-type energy storage, considering the coupling characteristics of system frequency dynamics and energy storage dynamics, and optimize the optimal control strategy under the objective of minimizing the cost of grid-type energy storage; 4) Analyze the linearized constraints of the optimal control framework to achieve linearized representation of system frequency dynamics and energy storage dynamics; 5) The analytical model predicts the linearized objective function of the optimal control framework, realizing the linear mapping between the optimization objective and the system state and energy storage state; 6) Construct a sequential linearization solution framework to transform the nonlinear optimization problem of the optimal control framework into a set of iterative linear optimization problems. Achieve fast and stable convergence of the algorithm through acceptance / rejection mechanisms and credibility region update mechanisms.

[0036] According to the method, step 1) specifically includes: The system frequency dynamic characteristic characterization model describes the system frequency response under virtual inertia and virtual damping provided by grid-based energy storage. (Grid node set) superscript Let each node of the power grid represent a node. Let the system state vector be... Including the power grid phase angle offset at each node Angular frequency offset of the power grid and synchronous generator set mechanical power offset ,Right now .

[0037] The system frequency dynamic characteristic characterization model has the following form:

[0038] in This represents the derivative of the system state vector with respect to time. This represents the control parameters for grid-based energy storage, including the virtual inertia of grid-based devices. Virtual damping ,Right now . This represents the load fluctuation vector at each node. It represents the derivative of the system state vector with respect to time.

[0039] and The system state matrix is ​​determined by the control variables of the grid-type energy storage system, specifically:

[0040] in Represents the identity matrix. Represents the zero matrix. These include the inertia of all nodes. and damping diagonal matrix, This is the susceptance matrix that includes the susceptance of all lines. The auxiliary parameter matrix is ​​defined as follows: ,in This includes the droop coefficient ratio of all nodes. and synchronous generator set time constant The diagonal matrix, for nodes not connected to synchronous generator sets, can be made into... , Each of these represents a non-zero value greater than 1. For nodes connected to synchronous generator sets, the mechanical power offset of the synchronous generator sets... , In the Laplace domain frequency offset The relationship is a downward proportional relationship:

[0041] System output Including the frequency offset of the center of inertia (COI) and its rate of change ,Right now The center of inertia offset is defined as the weighted average of the offsets of each node, weighted by its inertia. System output. The system state vector and power vector can be given by the system output equation as follows:

[0042] in and The system output matrix has the following structure:

[0043]

[0044] in The output matrix part corresponding to the inertia center frequency offset is calculated to obtain the frequency-weighted average value of each node with inertia as the weight. and This is the output matrix portion corresponding to the rate of change of the inertia center frequency offset. It is defined as follows:

[0045]

[0046]

[0047] in This is an inertia vector that includes the inertia of each node. This is the sum of the inertia of all nodes.

[0048] When the total number of system nodes When the value is greater than 2, all nodes except for the grid-connected energy storage nodes can be equivalently aggregated into a single synchronous generator unit, thus simplifying the entire power grid into a single-unit infinite bus system. Let the nodes excluding the grid-connected energy storage nodes... The set of other nodes is The equivalent aggregation of a single synchronous generator set has the following equivalent parameters:

[0049] in For equivalent inertia, For equivalent damping, This is the equivalent droop coefficient ratio. The time constant is the equivalent synchronous generator set. The system frequency dynamic characteristic characterization model degenerates into a model describing a two-node system, where one node corresponds to grid-type energy storage and the other node corresponds to an equivalent aggregated synchronous generator set, and has the aforementioned equivalent parameters.

[0050] Output power of grid-type energy storage This can be obtained from the power flow in the aforementioned two-node system. Based on DC power flow, It should be:

[0051] in The susceptance matrix that aggregates all line susceptance corresponds to the node. Self-susceptivity, The phase angle of synchronous generator units, including grid-type energy storage and equivalent aggregation. For nodes Load fluctuations at the location.

[0052] The continuous state space is discretized using the backward Euler method with zero-order preservation, yielding the frequency response model in the discrete state space. (Parameter control quantity) The decision is obtained at each control point and remains constant across all sampling points within that control point. ,in For the first Parameter control quantities at each control point For the first The parameter control quantity at each sampling point. The frequency response model in discrete state space can be expressed as:

[0053] The first in discrete state space System state matrix at each control point Given the system state matrix in continuous state space:

[0054]

[0055] Using the system state in discrete state space The system output in discrete state space can be obtained. The system output equation remains unchanged.

[0056] According to the method, step 2) specifically includes: The energy storage dynamic characteristic characterization model describes the state update relationship during the dynamic operation of grid-type energy storage. Let the energy storage state vector... Including state of charge and polarization voltage ,Right now .

[0057] The energy storage dynamic characteristic characterization model has the following form:

[0058] in This represents the derivative of the energy storage state vector with respect to time. This indicates the energy storage input current.

[0059] and The time-invariant state transition matrix is ​​determined by the parameters of the grid-type energy storage cells, specifically:

[0060] in These are the cell polarization resistor and capacitor, respectively. This represents the energy storage capacity. There is a non-linear mapping relationship between the open-circuit voltage and the state of charge of the energy storage cell, which can be expressed as: Let the ohmic resistance of the battery cell be... The output of the energy storage dynamic characteristic characterization model is the cell port voltage. :

[0061] There is an inverter efficiency loss between the output power and the internal cell power of grid-connected energy storage, resulting in the inverter's charging and discharging efficiencies being respectively... and The number of cells in a grid-type energy storage system is The output power of grid-type energy storage and internal cell power The relationship is:

[0062] in Indicates discharge. Indicates charging. Because... Characterized by the system's frequency dynamics, power can be used as a variable to solve for the power of a given internal cell. Current of under-grid energy storage:

[0063] The continuous state space is discretized using the backward Euler method with zero-order preservation, yielding the frequency response model in the discrete state space. The dynamic characteristic model of energy storage in the discrete state space can be expressed as:

[0064] The system state matrix in discrete state space The system state matrix in continuous state space Given:

[0065]

[0066] Using the system state in discrete state space The system output in discrete state space can be obtained. The system output equation remains unchanged.

[0067] According to the method described above, step 3) specifically includes: The model prediction optimal control framework for grid-based energy storage is based on the control quantity of grid-based energy storage. This is an optimization problem with energy storage cost as the decision variable and the feasible region of energy storage operation as the constraint.

[0068] To ensure the robustness of grid-type energy storage control under varying degrees of disturbance, this optimal control framework should provide control commands under all disturbance levels. Therefore, frequency stability requirements are incorporated into the objective function. By penalizing deviations exceeding the frequency stability requirement, the frequency offset is kept as close to the stability requirement as possible through a penalty function. This design allows the control framework to still provide usable control commands under large disturbances. If the frequency stability requirement is treated as a hard constraint, there might be unsolvable optimization problems under large disturbances, resulting in the inability to provide usable control commands. The deviations requiring penalty are defined as follows:

[0069] in This is the upper limit of frequency offset. This represents the upper limit of the frequency change rate offset. For frequency offset exceeding the limit, For frequency change rate deviation exceeding the limit, This indicates taking the larger of the two values.

[0070] Define the cost of grid-based energy storage as Its physical meaning is the energy arbitrage opportunity lost by energy storage when connected to the grid through a grid-connected inverter.

[0071]

[0072] in This represents the arbitrage profit per unit kWh of energy under the optimal day-ahead scheduling scheme.

[0073] The optimization problem of the model prediction optimal control framework for the grid-type energy storage is as follows:

[0074] In the formula Describe the objective function. To exceed the limit penalty coefficient, These are the upper and lower limits of the permissible state of charge (SOC) for grid-type energy storage operation, respectively. These are the upper and lower limits of the permissible current during the operation of grid-type energy storage. This represents the rated feasible range for grid-type energy storage control.

[0075] According to the method described above, step 4) specifically includes: The aforementioned network-type energy storage model predicts an optimal control framework that exhibits high nonlinearity. The linearized constraints enable a linearized characterization of both system frequency dynamics and energy storage dynamics.

[0076] Introducing auxiliary control quantities . With parameter control quantity There is a one-to-one correspondence, therefore, obtaining the pair The optimization problem can be equivalent to obtaining the... The optimization problem involves defining the timing aggregation of control variables at all control points. vector Let the original optimal solution be... The system frequency dynamic characteristic characterization model in By performing a Taylor expansion in the vicinity, the linearized frequency dynamics of the system can be obtained.

[0077]

[0078] In the formula This represents the improved and optimized solution. Batch state matrix for time-series aggregation: , Batch output matrix for time-series aggregation: , The optimal solution is respectively represented as The batch state matrix and batch output matrix of the time-series aggregation are shown below. The optimal solution is respectively represented as The batch state gradient and batch output gradient are given.

[0079] The slope of the open-circuit voltage curve of the energy storage cell is defined as follows: Since the open-circuit voltage curve of lithium iron phosphate material commonly used in energy storage cells is relatively smooth, it can be assumed that the slope of the open-circuit voltage under the improved optimization solution is equal to the slope under the original optimization solution, that is:

[0080] The open-circuit voltage of the energy storage cell can be linearized as follows:

[0081] in Indicates the first The sampling point and the first The change in the state of charge between sampling points. Substituting this into the above equation, the equation becomes: and Linear constraints as decision variables.

[0082] Let the charging and discharging currents of the grid-type energy storage be respectively and All of these are decision variables. In improving the optimization solution... The constraints used to differentiate between energy storage charging and discharging processes include:

[0083] To distinguish the charging and discharging processes of energy storage without introducing additional 0 and 1 variables, and to calculate the conversion between the internal cell power and output power of grid-type energy storage based on the charging and discharging direction, an equivalent current from the grid side is introduced. As an auxiliary variable:

[0084] Grid-type energy storage output power Equivalent current from the grid side can be utilized The calculation yielded: .

[0085] Let the auxiliary voltage At this time, the output voltage of the energy storage cell is In the original optimal solution nearby, right An improved optimal solution can be obtained by performing a Taylor expansion. Output power of grid-type energy storage with down-linearization:

[0086] In the formula For the original optimal solution Given a constant, Since the variables to be optimized are decision variables, the physical essence of the above equation is an equality constraint.

[0087] According to the method, step 5) specifically includes: The linearized objective function enables a linear mapping between the optimization objective and the system state and energy storage state.

[0088] The process of truncating the portion of the excess quantity that exceeds the limit in the objective function is nonlinear. A nonnegative auxiliary decision variable is introduced. Replace the original problem The aforementioned auxiliary decision variables effectively replace the process of truncating non-negative out-of-bounds values ​​when the following constraints are met:

[0089] The process of taking the absolute value of the grid-type energy storage cost in the objective function is nonlinear. An auxiliary decision variable is introduced. When the auxiliary decision variable satisfies the following constraints, it is equivalent to replacing the process of taking the absolute value:

[0090]

[0091]

[0092] Cost of linear grid-based energy storage for: .

[0093] According to the method, step 6) specifically includes: Based on the linearization constraints and linearization objective function in steps 4) and 5), the original optimization problem can be solved as follows: The problem can be rewritten as the following linearization optimization problem:

[0094] in Represents the linearization objective function. Let be the decision variables to be optimized. express Nearby trusted domains, Indicates the size of the trust region, then It should be:

[0095] The sequence linearization solution framework includes an iterative loop of linearization → solution → update: Let the current iteration number be... From the initial solution Starting from the current optimal solution The original linearization problem is solved nearby, and the result is updated. Then increment the iteration count by 1 and repeat the above steps.

[0096] The aforementioned acceptance / rejection mechanism refers to the mechanism that, when the above-mentioned information is obtained... After solving the linearization optimization problem in the vicinity, solve the above problem. If the linearization optimization problem has an optimal solution... Then evaluate the descent of its linearized objective function. and the decrease of the true objective function :

[0097]

[0098] in Indicates up to the The optimal solution obtained in the next iteration. Let the minimum acceptance rate be... Accept when the following conditions are met As the new optimal solution ,and Updated to , :

[0099] Otherwise, it should be refused. As the new optimal solution However, to give the algorithm a certain degree of randomness and thus avoid getting trapped in local optima, weights are used... Update the optimal solution :

[0100] The trusted domain update mechanism refers to when When the solution is accepted as the new optimal solution, it indicates that the linearization is effective. The nearby problem structure has a low degree of nonlinearity, and the credibility region can be determined according to the acceptance factor. Increase the size appropriately to improve the convergence speed:

[0101] in Indicates the first The size of the confidence region to be used in the next iteration This represents the scaling factor of the confidence region size relative to the current solution. These represent the upper and lower bounds of the trust domain size, respectively. The trust domain is updated accordingly based on the updated trust domain size.

[0102] when When a solution is rejected as a new optimal solution, it indicates that the linearization process is ineffective. The nearby problem structure has a high degree of nonlinearity, and the confidence region can be calculated based on the rejection factor. Reduce the size appropriately and solve for linearization again:

[0103] The trusted domain is updated accordingly based on the updated trusted domain size:

[0104] In the In the next iteration, if the above linearized optimization problem has no solution, the following linearized relaxation problem can be solved to obtain the optimal direction to return to the feasible region:

[0105] in As slack variables, Denotes the first-order norm. This represents the penalty coefficient for the corresponding slack variable. Solving the above relaxation problem yields the optimal solution. ,renew : But the optimal solution No update. Meanwhile, the size of the trusted domain depends on... Update:

[0106] The trusted domain is updated accordingly based on the updated trusted domain size:

[0107] The above iterative process stops when the algorithm converges. The step size convergence threshold is defined as... The convergence threshold of the objective function is To maintain the stability of the algorithm's convergence, the algorithm terminates when the convergence condition is met for three consecutive iterations. Let the set of the three consecutive iterations be... , Let be the current iteration number. The convergence condition is:

[0108] The above sequence linearization solution framework can be summarized in Table 1 below.

[0109] Table 1

[0110] To achieve the above embodiments, such as Figure 2 As shown, this embodiment also provides a sequence linearization solution device 10 for the optimal control problem of grid-type energy storage. The device 10 includes a system frequency modeling module 100, an energy storage dynamic modeling module 200, an optimization framework construction module 300, a linearization processing module 400, and a sequence solution module 500.

[0111] The system frequency modeling module 100 is used to establish a system frequency dynamic characteristic characterization model to depict the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. The Energy Storage Dynamic Modeling Module 200 is used to establish a dynamic characteristic characterization model of energy storage, and to depict the state changes and output changes of grid-type energy storage during dynamic operation. The optimization framework construction module 300 is used to construct the model prediction optimal control framework for grid-type energy storage based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The linearization processing module 400 is used to analyze the linearization constraints of the optimal control framework predicted by the analytical model, and realize the linearization representation of the system frequency dynamics and energy storage dynamics; it performs linearization processing on the linearization objective function of the optimal control framework predicted by the analytical model, and realizes the linear mapping between the optimization objective and the system state and energy storage state. The sequence solving module 500 is used to execute the sequence linearization solution framework. It solves the original nonlinear optimization problem by iteratively solving the linear optimization subproblems and supports acceptance / rejection mechanisms and credibility region update mechanisms.

[0112] Furthermore, the aforementioned system frequency modeling module 100 is also used to construct a system frequency dynamic characteristic characterization model based on discrete state space, adopting a model predictive control framework, defining the sampling step size, control step size and prediction interval, and discretizing the continuous state space model through the backward Euler method. The system frequency dynamic characteristic characterization model has the following form:

[0113] in This represents the derivative of the system state vector with respect to time. and The system state matrix is ​​determined by the control variables of the grid-type energy storage. This represents the control parameters for grid-type energy storage, including the virtual inertia and virtual damping of grid-type devices. This represents the load fluctuation vector at each node. The system state vector includes the grid phase angle offset, grid angular frequency offset, and synchronous generator mechanical power offset at each node in the grid node set.

[0114] Furthermore, the aforementioned energy storage dynamic modeling module 200 is also used to characterize the energy storage dynamic characteristics model, including the state update relationship between the energy storage state of charge and polarization voltage, and to consider the nonlinear mapping relationship between the open-circuit voltage of the energy storage cell and the state of charge, as well as the impact of inverter efficiency loss on output power. The energy storage dynamic characteristic characterization model has the following form:

[0115] in This represents the derivative of the energy storage state vector with respect to time. Indicates the energy storage input current. and The time-invariant state transition matrix is ​​determined by the parameters of the grid-type energy storage cells. This represents the energy storage state vector, which includes the state of charge and polarization voltage.

[0116] Furthermore, the aforementioned optimization framework construction module 300 is also used to incorporate the frequency stability requirement into the objective function in the form of a penalty function in the model prediction optimal control framework, so as to give feasible control commands under large disturbances by penalizing the frequency offset exceeding the limit. The optimization problem of the model prediction optimal control framework for the grid-type energy storage is as follows:

[0117] In the formula Describe the objective function. To exceed the limit penalty coefficient, These are the upper and lower limits of the permissible state of charge (SOC) for grid-type energy storage operation, respectively. These are the upper and lower limits of the permissible current during the operation of grid-type energy storage. This represents the rated feasible range for grid-type energy storage control.

[0118] Furthermore, the linearization processing module 400 described above is also used for: The optimization variables are reconstructed by introducing auxiliary control variables; Taylor expansion is performed on the system frequency dynamic characteristic characterization model near the original optimization solution to obtain the linearized system state and output relationship; The open-circuit voltage curve of the energy storage cell is approximated by a piecewise linear approach; the equivalent current on the grid side is introduced as an auxiliary variable to linearize the coupling relationship between the output power of the grid-type energy storage and the power of the internal cells. Introduce a non-negative auxiliary variable to equivalently replace the truncation treatment that exceeds the frequency limit; By introducing an auxiliary variable to replace the calculation of the absolute value of power, the energy storage cost can be expressed in a linear form.

[0119] Furthermore, the sequence solving module 500 described above is also used for: The original problem is linearized near the current optimal solution to obtain a linear optimization subproblem; this linear optimization subproblem is solved, and the acceptance / rejection mechanism is used to determine whether to accept the new solution; the confidence region size is dynamically updated based on the acceptance or rejection result; if the linear optimization subproblem has no feasible solution, the linearized relaxation problem with relaxation variables is solved to obtain the search direction of the regression feasible region; The acceptance / rejection mechanism determines whether to accept a new solution by comparing the decrease value of the linearized objective function with the decrease value of the actual objective function and combining it with a preset minimum acceptance rate; the trust region update mechanism expands or shrinks the trust region range proportionally based on the acceptance status of the current solution.

[0120] This invention provides a sequential linearization solution device for the optimal control problem of network-type energy storage, which can construct a unified latent space that simultaneously supports visual generation and perception understanding tasks, maintain the semantic structure of the self-supervised encoder and improve reconstruction quality, significantly reduce training and deployment costs, and enhance the transferability and scalability of the model.

[0121] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 3 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the method described above.

[0122] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0123] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0124] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A sequential linearization solution method for the optimal control problem of grid-type energy storage, characterized in that, include: A system frequency dynamic characteristic characterization model is established to characterize the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. Establish a dynamic characteristic characterization model for energy storage to depict the state changes and output changes of grid-type energy storage during dynamic operation; Based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model, a model prediction optimal control framework for grid-type energy storage is constructed. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The analytical model predicts the linearized constraints of the optimal control framework, realizing the linearized characterization of system frequency dynamics and energy storage dynamics; The linearized objective function of the analytical model predicting the optimal control framework is linearized to achieve a linear mapping between the optimization objective and the system state and energy storage state; A sequential linearization solution framework is constructed, which transforms the nonlinear optimization problem of the model predicting the optimal control framework into a set of iterative linear optimization problems for solution. The algorithm achieves fast and stable convergence through acceptance / rejection mechanisms and a credibility region update mechanism.

2. The method as described in claim 1, characterized in that, The linearization process of the constraints includes: The optimization variables are reconstructed by introducing auxiliary control variables; Taylor expansion is performed on the system frequency dynamic characteristic characterization model near the original optimization solution to obtain the linearized system state and output relationship; The open-circuit voltage curve of the energy storage cell is approximated by a piecewise linear approach; the equivalent current on the grid side is introduced as an auxiliary variable to linearize the coupling relationship between the output power of the grid-type energy storage and the power of the internal cells. The equivalent current on the grid side is: , Grid-type energy storage output power Equivalent current from the grid side can be utilized The calculation yielded: ; In the original optimal solution nearby, right The improved optimal solution is obtained by performing a Taylor expansion. Output power of grid-type energy storage with down-linearization: In the formula For the original optimal solution Given a constant, These are the decision variables to be optimized.

3. The method as described in claim 1, characterized in that, The linearization of the objective function includes: Introduce a non-negative auxiliary variable to equivalently replace the truncation treatment that exceeds the frequency limit; By introducing an auxiliary variable to replace the calculation of the absolute value of power, the energy storage cost can be expressed in a linear form.

4. The method as described in claim 1, characterized in that, The iterative process of the sequence linearization solution framework includes: The original problem is linearized near the current optimal solution to obtain a linear optimization subproblem; this linear optimization subproblem is solved, and the acceptance / rejection mechanism is used to determine whether to accept the new solution; the confidence region size is dynamically updated based on the acceptance or rejection result; if the linear optimization subproblem has no feasible solution, the linearized relaxation problem with relaxation variables is solved to obtain the search direction of the regression feasible region; The acceptance / rejection mechanism determines whether to accept a new solution by comparing the decrease value of the linearized objective function with the decrease value of the actual objective function and combining it with a preset minimum acceptance rate; the trust region update mechanism expands or shrinks the trust region range proportionally based on the acceptance status of the current solution.

5. The method as described in claim 4, characterized in that, The acceptance / rejection mechanism is as follows: when acquiring in If a linearized optimization problem in the vicinity has an optimal solution, then... Then evaluate the descent of its linearized objective function. and the decrease of the true objective function : in Indicates up to the The optimal solution obtained in the next iteration has a minimum acceptance rate. When satisfied ,accept As the new optimal solution ,and Updated to , ; Otherwise, it should be refused. As the new optimal solution And by weight Update the optimal solution : ； The trusted domain update mechanism is as follows: when When the solution is accepted as the new optimal solution, it indicates that the linearization is effective. The nearby problem structure has a low degree of nonlinearity, and the credibility region can be determined according to the acceptance factor. Increase the size appropriately to improve the convergence speed. ， in Indicates the first The size of the confidence region to be used in the next iteration This represents the scaling factor of the confidence region size relative to the current solution. These represent the upper and lower bounds of the trust domain size, respectively; the trust domain is updated accordingly based on the updated trust domain size. ； when When a solution is rejected as a new optimal solution, it indicates that the linearization process is ineffective. The nearby problem structure has a high degree of nonlinearity, and the confidence region can be calculated based on the rejection factor. Reduce the size appropriately and solve for linearization again: ， The trusted domain is updated accordingly based on the updated trusted domain size: 。 6. The method as described in claim 1, characterized in that, The established system frequency dynamic characteristic characterization model is based on discrete state space, adopts model predictive control framework, defines sampling step size, control step size and prediction interval, and discretizes continuous state space model through backward Euler method; The system frequency dynamic characteristic characterization model has the following form: in This represents the derivative of the system state vector with respect to time. and The system state matrix is ​​determined by the control variables of the grid-type energy storage. This represents the control parameters for grid-type energy storage, including the virtual inertia and virtual damping of grid-type devices. This represents the load fluctuation vector at each node. The system state vector includes the grid phase angle offset, grid angular frequency offset, and synchronous generator mechanical power offset at each node in the grid node set.

7. The method as described in claim 1, characterized in that, The energy storage dynamic characteristic characterization model includes the state update relationship between the energy storage state of charge and polarization voltage, and considers the nonlinear mapping relationship between the open circuit voltage of the energy storage cell and the state of charge, as well as the impact of inverter efficiency loss on output power. The energy storage dynamic characteristic characterization model has the following form: in This represents the derivative of the energy storage state vector with respect to time. Indicates the energy storage input current. and The time-invariant state transition matrix is ​​determined by the parameters of the grid-type energy storage cells. This represents the energy storage state vector, which includes the state of charge and polarization voltage.

8. The method as described in claim 1, characterized in that, The model predicts the optimal control framework by incorporating the frequency stability requirement into the objective function in the form of a penalty function. By penalizing the frequency offset exceeding the limit, it can still provide feasible control commands under large disturbances. The optimization problem of the model prediction optimal control framework for the grid-type energy storage is as follows: In the formula Describe the objective function. To exceed the limit penalty coefficient, These are the upper and lower limits of the permissible state of charge (SOC) for grid-type energy storage operation, respectively. These are the upper and lower limits of the permissible current during the operation of grid-type energy storage. This represents the rated feasible range for grid-type energy storage control.

9. A sequential linearization solution device for the optimal control problem of grid-type energy storage, characterized in that, include: The system frequency modeling module is used to establish a dynamic characteristic model of the system frequency, and to characterize the system frequency response under real-time disturbances and when grid-type energy storage provides virtual inertia and virtual damping. The energy storage dynamic modeling module is used to establish a dynamic characteristic model of energy storage and to characterize the state changes and output changes of grid-type energy storage during dynamic operation. The optimization framework construction module is used to construct the model prediction optimal control framework for grid-type energy storage based on the system frequency dynamic characteristic characterization model and the energy storage dynamic characteristic characterization model. The grid-type energy storage control quantity is used as the decision variable, the energy storage cost is minimized as the objective, the energy storage operation feasible domain is used as the constraint condition, and the coupling characteristics of system frequency dynamics and energy storage dynamics are considered. The linearization module is used to analyze the linearized constraints of the model predicting the optimal control framework, and realize the linearized representation of the system frequency dynamics and energy storage dynamics; it analyzes the linearized objective function of the model predicting the optimal control framework and performs linearization processing to realize the linear mapping between the optimization objective and the system state and energy storage state. The sequence solving module is used to execute the sequence linearization solution framework. It solves the original nonlinear optimization problem by iteratively solving the linear optimization subproblems, and supports acceptance / rejection mechanisms and credibility region update mechanisms.

10. A computer device, characterized in that, Including processor and memory; The processor reads the executable program code stored in the memory to run the program corresponding to the executable program code, so as to implement the sequential linearization solution method for the optimal control problem of grid-type energy storage as described in any one of claims 1-8.

Images