Power system scheduling method and device based on multi-dimensional full-pure embedded approximate dynamic programming, terminal equipment and storage medium

Abstract

The invention discloses an electric power system scheduling method and device based on multi-dimensional full-pure embedding approximate dynamic programming, terminal equipment and a storage medium, and belongs to the field of electric power systems.The method comprises the steps that optimization constraints are constructed, and an optimization model represented by a plurality of decoupled value functions is constructed; converting each decoupled value function into a corresponding Lagrange function, and then solving a gradient to obtain a plurality of KKT conditions; embedding preset complex parameters into each KKT condition to obtain a plurality of polynomial equation sets; calculating a zero-order coefficient of each polynomial equation set, and performing recursive calculation on a non-zero-order coefficient of each polynomial equation set; and obtaining an approximate expression of a value function corresponding to each polynomial equation set according to the obtained coefficient solution of each order, obtaining a solution of an optimization model according to the approximate expression, and scheduling the power system. By implementing the method and the device, the defects of low efficiency and poor effect in solving the scheduling problem of the power system in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the field of power system dispatching, and in particular to a power system dispatching method, device, terminal equipment and storage medium based on multi-dimensional holomorphic embedding approximate dynamic programming. Background Art

[0002] Approximate dynamic programming, as an emerging algorithm, has been successfully applied in fields such as power optimization and scheduling in recent years. Within the framework of approximate dynamic programming, random multi-period scheduling problems can be decomposed into a series of solvable single-period subproblems, with value functions representing the interactions between the subproblems. Since value functions do not have explicit expressions, the core issue of approximate dynamic programming lies in how to approximate them. Among existing approximate dynamic programming algorithms, value function approximation based on piecewise linear functions and value function approximation based on cuts can, to a certain extent, compensate for the shortcomings of table-based value function approximation techniques. However, by only using first-order functions for approximation, they cannot effectively solve the nonlinear problems of power system scheduling. Furthermore, these traditional value function approximation techniques require iterative updates of the approximate value function until convergence to the actual value function, reducing algorithm efficiency. Summary of the Invention

[0003] The present invention provides a power system scheduling method, apparatus, terminal equipment and storage medium based on multidimensional holomorphic embedding approximate dynamic programming. The method effectively overcomes the defects of low efficiency and poor effect of existing value function approximation technology in solving power system scheduling problems.

[0004] An embodiment of the present invention provides a power system scheduling method based on multidimensional holomorphic embedding approximate dynamic programming, comprising: Obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and dispatch cost coefficients; Based on the acquired data, optimization constraints are established, and with the goal of minimizing the total cost of the power system, an optimization model represented by a plurality of decoupled value functions is constructed; wherein each of the value functions is used to represent the optimization problem corresponding to a power operation scenario in the power operation scenario set within a predetermined sub-period to be optimized; Based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated to obtain the KKT condition corresponding to each Lagrangian function; Embedding a preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded with the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group; For each of the polynomial equations, recursively calculating the non-zero-order coefficients of the polynomial equations according to the zero-order coefficients until the solved order of the polynomial equations reaches a preset maximum order, terminating the recursive calculation operation, obtaining coefficient solutions of each order of the polynomial equations, and obtaining an approximate expression of the value function corresponding to the polynomial equations according to the coefficient solutions; The solution of the optimization model is solved according to the approximate expression of each value function, and the power system is dispatched according to the solution of the optimization model.

[0005] Further, the complex parameters include a first complex parameter; The embedding of the preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the first complex parameter into each KKT condition, and expanding the variables in the KKT condition embedded in the first complex parameter in a power series form to obtain a first polynomial equation group corresponding to each KKT condition; For each of the first polynomial equation groups, merge like terms of the first polynomial equation group to obtain a first polynomial equation group after merging like terms; extract coefficients of each like term in the first polynomial equation group after merging like terms; construct a coefficient equation group based on preset coefficients and conditions and the coefficients of each like term; solve the coefficients of zero-order like terms in the coefficient equation group to determine the zero-order coefficient of the first polynomial equation group; The zero-order coefficients of each of the first polynomial equations are obtained.

[0006] Further, the complex parameter includes a second complex parameter; the optimization constraint includes a first operating equation constraint; the first operating equation constraint includes an active power balance constraint and a reactive power balance constraint; the KKT condition includes a first Lagrangian multiplier corresponding to the active power balance constraint and a second Lagrangian multiplier corresponding to the reactive power balance constraint; The method of embedding a preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the second complex parameter into each KKT condition, and decoupling the equation with node coupling in the KKT condition embedded in the second complex parameter based on a preset relaxation parameter to obtain a decoupled KKT condition; Expanding the variables in each of the decoupled KKT conditions in a power series form to obtain a corresponding second polynomial equation group; assigning specific values to the second complex parameters; Assigning the same value to the first Lagrangian multiplier and the second Lagrangian multiplier; According to the second complex parameter after being assigned a value, the first Lagrange multiplier after being assigned a value, the second Lagrange multiplier after being assigned a value and each of the second polynomial equation groups, the zero-order coefficient corresponding to each of the second polynomial equation groups is obtained by solving.

[0007] Furthermore, the recursive calculation operation includes: The order coefficient of the current order to be solved is calculated based on the order coefficient corresponding to the solved order; wherein, when the current order to be solved is first order, the order coefficient corresponding to the solved order is the zero-order coefficient; Determine whether the current order to be solved reaches the maximum order, If not, the current order to be solved is used as the solved order, and the next order after the solved order is used as the current order to be solved for the next recursive calculation operation; If so, terminate the recursive calculation operation.

[0008] Furthermore, each of the power operation scenarios includes a set of state space parameters and a set of state transition parameters; the parameters to be optimized include the output of the thermal power unit, the charge and discharge power of the energy storage power station in each node, the real and imaginary components of the voltage vector of each node, and the transmission power used for power interaction with the upper power grid; The optimization constraints are constructed based on the acquired data, and with the goal of minimizing the total cost of the power system, an optimization model represented by several decoupled value functions is constructed, including: Based on the acquired data, optimization constraints are established, and an initial optimization model is constructed using the output of the thermal power units, the charge and discharge power of the energy storage power station at each node, the real and imaginary components of the voltage vector at each node, and the transmission power for power interaction with the upper power grid as decision variables, with the goal of minimizing the total cost of the power system; According to a preset approximate dynamic programming algorithm, the initial optimization model is decoupled to obtain the optimization model represented by the plurality of decoupled value functions.

[0009] Furthermore, the expression of the decoupled value function is specifically: in, is the value function of the (t-Δt)th sub-period to be optimized in the nth power operation scenario after decoupling, is the state space parameter of the (t-Δt)th sub-period to be optimized; is the instantaneous profit function, which represents the profit of the (t-Δt)th sub-period to be optimized in the nth power operation scenario, x tis the parameter to be optimized in the tth sub-period to be optimized, is the state transition parameter of the nth power operation scenario in the tth sub-period to be optimized; is the value function obtained by Monte Carlo method, is the state space parameter of the tth sub-period to be optimized after the optimization decision is made and before the state transfer parameter is updated; T is the set of sub-periods to be optimized, which consists of all sub-periods to be optimized; N W Run scenario sets for electricity.

[0010] Furthermore, the optimization constraint further includes: running inequality constraints; Based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated respectively to obtain the KKT conditions corresponding to each Lagrangian function, including: Introducing a preset slack variable into the running inequality constraint so that the running inequality constraint is converted into an equation form to obtain a second running equality constraint; Determine the lower limit expression of the slack variable according to the preset barrier factor and the preset slack variable value constraint; Based on the first operating equality constraint, the second operating equality constraint, and the slack variable lower limit expression, converting each decoupled value function into a corresponding Lagrangian function; The gradient of the Lagrangian function is calculated to obtain the KKT conditions corresponding to each Lagrangian function.

[0011] An embodiment of the present invention further provides a power system dispatching device based on multidimensional holomorphic embedding approximate dynamic programming, comprising: a data acquisition module, an optimization model construction module, a KKT condition determination module, a KKT condition processing module, a value function approximate expression solving module, and a dispatching module; The data acquisition module is used to obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and a dispatch cost coefficient; The optimization model construction module is used to construct optimization constraints based on the acquired data, and to construct an optimization model represented by a plurality of decoupled value functions with the goal of minimizing the total cost of the power system; wherein each of the value functions is used to represent the optimization problem corresponding to a certain power operation scenario in the power operation scenario set within a preset sub-period to be optimized; The KKT condition determination module is used to convert each decoupled value function into a corresponding Lagrangian function based on the optimization constraints, and calculate the gradient of each Lagrangian function to obtain the KKT condition corresponding to each Lagrangian function; The KKT condition processing module is used to embed a preset complex parameter into each KKT condition, expand the variables in the KKT condition embedded with the complex parameter in the form of a power series to obtain a polynomial equation group corresponding to each KKT condition, and calculate the zero-order coefficient of each polynomial equation group; The value function approximate expression solving module is configured to perform a recursive calculation operation on the non-zero-order coefficients of each polynomial equation group according to the zero-order coefficient, until the solved order of the polynomial equation group reaches a preset maximum order, terminate the recursive calculation operation, obtain coefficient solutions of each order of the polynomial equation group, and obtain an approximate expression of the value function corresponding to the polynomial equation group according to the coefficient solutions; The scheduling module is used to solve the solution of the optimization model based on the approximate expressions of each value function, and to schedule the power system based on the solution of the optimization model.

[0012] The present application also provides a terminal device, including: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming as described in the above-mentioned embodiment of the invention.

[0013] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming as described in the above-mentioned embodiment of the invention is implemented.

[0014] The following beneficial effects are achieved by implementing the present invention: The present invention provides a method, apparatus, terminal device, and storage medium for power system scheduling based on multidimensional holomorphic embedded approximate dynamic programming. The method first constructs optimization constraints based on acquired data, and with the goal of minimizing the total cost of the power system, constructs an optimization model represented by several decoupled value functions. This decomposes the nonlinear problem of power system scheduling into multiple small-scale sub-problems. Then, based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated respectively to obtain the KKT condition corresponding to each Lagrangian function. A preset complex parameter is embedded in each KKT condition, and the variables in the KKT condition embedded with the complex parameter are expanded in the form of a power series to obtain a polynomial equation group corresponding to each KKT condition. Thus, by embedding the complex parameter, the value function is expanded in the form of a power series. The polynomial equation group obtained after the expansion can more accurately handle the nonlinear characteristics of the power system scheduling problem, avoiding the limitation of the traditional approximation method that only uses first-order function approximation. Calculate the zero-order coefficients of each polynomial equation group, and based on the zero-order coefficients, perform a recursive calculation operation on the non-zero-order coefficients of the polynomial equation group until the solved order of the polynomial equation group reaches a preset maximum order, terminate the recursive calculation operation, obtain the coefficient solutions of each order of the polynomial equation group, and obtain an approximate expression of the value function corresponding to the polynomial equation group based on the coefficient solutions; thus, the present application extends the value function from a first-order function to a higher-order function by expanding the order of the approximate value function, effectively avoiding the iterative update of the value function in the prior art and improving computational efficiency; Finally, based on the approximate expressions of each value function, the solution of the optimization model is solved, and the power system is dispatched according to the solution of the optimization model; thus, by solving the solution of the optimization model based on the approximate expressions of each value function, the total cost of the power system can be effectively reduced while ensuring the safe and stable operation of the power system, and a better dispatching goal can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 This is a flowchart of a power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming provided by a certain embodiment of the present application; Figure 2 This is a schematic diagram of the optimization model solution process provided by a certain embodiment of the present application; Figure 3 This is a structural diagram of a power system dispatching device based on multidimensional holomorphic embedding approximate dynamic programming provided by a certain embodiment of the present application; Figure 4 This is a schematic diagram of the structure of a terminal device provided in a certain embodiment of the present application. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0019] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0020] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0021] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0022] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0023] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0024] See also Figure 1 To address the low efficiency and poor performance of existing value function approximation technologies in solving power system scheduling problems, an embodiment of the present invention provides a power system scheduling method based on multidimensional holomorphic embedding approximate dynamic programming, comprising: S1. Obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and a dispatch cost coefficient; Specifically, obtain the power operation scenario set N W , parameters to be optimized, a set of power equipment, a set of nodes N, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters and a dispatching cost coefficient; wherein the power equipment set includes a set of conventional thermal power units D, an energy storage set E, a set of wind farms W and a set of photovoltaic power stations P; specifically, the dispatching cost coefficient includes a cost coefficient of conventional thermal power units, a cost coefficient of energy storage, a carbon trading price, a carbon emission allocation per unit of electricity, a coal consumption coefficient of conventional thermal power units, a carbon conversion coefficient, a wind farm's wind and solar power abandonment cost coefficient, a photovoltaic power station's wind and solar power abandonment cost coefficient, and a load abandonment cost coefficient; Specifically, the operating limits of the power equipment include the upper and lower limits of the output of conventional thermal power units, the upper limit of the charge and discharge power of the energy storage, the minimum value of the remaining power of the energy storage, the maximum value of the remaining power of the energy storage, the minimum value of the voltage amplitude of each node, and the maximum value of the voltage amplitude of each node; Specifically, the inter-node line admittance parameter includes the real part of the inter-node line admittance; the imaginary part of the inter-node line admittance; the real part of the node voltage vector; the imaginary part of the node voltage vector; Specifically, the grid interaction power limit includes a minimum value of active power exchanged with the upper grid, a maximum value of active power exchanged with the upper grid, a minimum value of reactive power exchanged with the upper grid, and a maximum value of reactive power exchanged with the upper grid.

[0025] S2. Constructing optimization constraints based on the acquired data and, with minimizing the total cost of the power system as the goal, constructing an optimization model represented by a plurality of decoupled value functions; wherein each value function is used to represent the optimization problem corresponding to a certain power operation scenario in the power operation scenario set within a predetermined sub-period to be optimized; In a preferred embodiment, each of the power operation scenarios includes a set of state space parameters and a set of state transition parameters; the parameters to be optimized include the output of the thermal power unit, the charging and discharging power of the energy storage power station in each node, the real and imaginary components of the voltage vector of each node, and the transmission power for power interaction with the upper power grid; The optimization constraints are constructed based on the acquired data, and with the goal of minimizing the total cost of the power system, an optimization model represented by several decoupled value functions is constructed, including: Based on the acquired data, optimization constraints are established, and an initial optimization model is constructed using the output of the thermal power units, the charge and discharge power of the energy storage power station at each node, the real and imaginary components of the voltage vector at each node, and the transmission power for power interaction with the upper power grid as decision variables, with the goal of minimizing the total cost of the power system; Decoupling the initial optimization model according to a preset approximate dynamic programming algorithm to obtain the optimization model represented by the plurality of decoupled value functions; Specifically, the parameter to be optimized represents the decision variable required by the initial optimization model to be constructed, and the parameter to be optimized x in the state of the tth sub-period to be optimized is i For example, the specific expression is as follows: Where, is the active power output of the conventional thermal power unit at node i in the tth sub-period to be optimized; is the reactive power output of the conventional thermal power unit at node i in the tth sub-period to be optimized; is the discharge power of the energy storage at node i in the tth sub-period to be optimized; is the charging power of the energy storage in the tth sub-period to be optimized at node i; e i，t is the real part of the voltage vector of node i in the tth sub-period to be optimized; f i，t The imaginary part of the voltage vector of node i in the tth sub-period to be optimized; P t gd is the active power exchanged with the upper power grid during the tth sub-period to be optimized; is the reactive power exchanged between node i and the upper power grid in the tth sub-period to be optimized; T is the set of sub-periods to be optimized, which consists of all sub-periods to be optimized.

[0026] Specifically, the state space parameter represents the state of the power system, and the state space parameter S of the tth sub-period to be optimized ist For example, the specific expression is as follows: Where, E i，t is the state of charge of the energy storage at node i in the tth sub-period to be optimized; is the active power output of the wind farm at node i in the tth sub-period to be optimized; is the active power output of the photovoltaic power station at node i in the tth sub-period to be optimized; Active load of node i in the tth sub-period to be optimized; is the reactive load of node i in the tth sub-period to be optimized; p t is the electricity price of the tth sub-period to be optimized; Specifically, the state transition parameter is used to characterize the random factor and is defined by the prediction error. The state transition parameter W of the tth sub-period to be optimized is t For example, the specific expression is as follows: Where, is the active power output of the wind farm in the tth sub-period to be optimized at node i The prediction error of is the active power output of the photovoltaic power station in the tth sub-period to be optimized at node i The prediction error of is the active load of node i in the tth sub-period to be optimized The prediction error of is the reactive load of node i in the tth sub-period to be optimized The prediction error of is the electricity price p of the tth sub-period to be optimized t The prediction error of Specifically, the transfer functions corresponding to the above prediction errors are as follows: Where W is the wind farm set; P is the photovoltaic power station set; is the prediction information of the active power output of the wind farm in the t+Δt-th sub-period to be optimized at node i; is the prediction information of the active power output of the photovoltaic power station in the t+Δtth sub-period to be optimized at node i, is the prediction information of active load of node i in the t+Δtth sub-period to be optimized, is the forecast information of reactive load of node i in the t+Δtth sub-period to be optimized; is the electricity price forecast information for the t+Δtth sub-period to be optimized.

[0027] Schematically, the constructed optimization constraints include an operating inequality constraint and a first operating equality constraint; the operating inequality constraint includes a power generation constraint, an energy storage constraint, a node voltage amplitude constraint, and an upper-level grid interaction power limit constraint; the first operating equality constraint includes an active power balance constraint, a reactive power balance constraint, and an energy storage state change constraint; Specifically, the specific calculation expression of power generation constraint is as follows: Where, is the lower limit of active power output of the conventional thermal power unit at node i; is the lower limit of reactive power output of conventional thermal power unit at node i; is the upper limit of active power output of conventional thermal power unit at node i; is the upper limit of reactive power output of conventional thermal power unit at node i; Specifically, the specific calculation expression of energy storage constraint is as follows: Where, E is the upper limit of charging and discharging power of energy storage at node i; i，min The minimum value of the remaining energy stored in node i; E i，max The maximum value of the remaining energy stored in node i; Specifically, the specific calculation expression of the node voltage amplitude constraint is as follows: Where V i，min is the minimum value of the voltage amplitude at node i; V i，max is the maximum voltage amplitude of node i; Specifically, the specific calculation expression of the upper-level grid interactive power limit constraint is as follows: Where, is the minimum value of active power exchanged with the upper grid; is the maximum value of active power exchanged with the upper grid; is the minimum value of reactive power exchanged with the upper grid; is the maximum value of reactive power exchanged with the upper power grid; specifically, the specific calculation expression of the active power balance constraint is as follows: Where G ij is the real part of the line admittance between node i and node j; B ij is the imaginary part of the line admittance between node i and node j; e j，t is the real part of the voltage vector of node j in the tth sub-period to be optimized; f j，tis the imaginary part of the voltage vector of node j in the tth sub-period to be optimized; Specifically, the specific calculation expression of reactive power balance constraint is as follows: Specifically, the specific calculation expression of the energy storage state change constraint is as follows: Where, E i，t+Δt is the state of charge of the energy storage at node i in the (t+Δt)th sub-period to be optimized.

[0028] Specifically, based on the state space parameters and state transition parameters in the power operation scenario set, the output of the thermal power units, the charge and discharge power of the energy storage power station in each node, the real and imaginary components of the voltage vector of each node, and the transmission power for power interaction with the upper power grid are used as decision variables, and the total cost of the power system is minimized. An initial optimization model is constructed, which is specifically expressed as follows: Where C t (S t , x t ) represents the benefit of making a scheduling decision under a certain state space parameter in the tth sub-period to be optimized, including the cost of conventional thermal power units Electricity purchase and sales charges Energy storage output cost Carbon trading costs and the cost of energy curtailment (including wind, solar and load curtailment) Specifically, the cost of conventional thermal power units is The specific calculation formula is as follows: Where, is the first cost coefficient of conventional thermal power unit at node i, is the second cost coefficient of conventional thermal power unit at node i, is the third cost coefficient of the conventional thermal power unit at node i, and Δt is the time step; The electricity purchase and sale charges The specific calculation formula is as follows: The energy storage output cost The specific calculation formula is as follows: Where, is the first cost coefficient of energy storage at node i, is the second cost coefficient of energy storage at node i, is the third cost coefficient of energy storage at node i, η i is the charging and discharging efficiency of energy storage at node i; The carbon trading costs The specific calculation formula is as follows: Where, is the carbon trading price; γ is the carbon emission quota per unit of electricity; μ i is the coal consumption coefficient of conventional thermal power unit at node i; ρ is the carbon conversion coefficient; The cost of curtailed electricity (wind, solar and load) The specific calculation formula is as follows: Where c wt is the wind farm’s curtailment cost coefficient; c pv is the cost coefficient of wind and solar curtailment in photovoltaic power stations; c load is the load abandonment cost coefficient.

[0029] Schematically, according to a preset approximate dynamic programming algorithm, the initial optimization model is decoupled to obtain the optimization model represented by several decoupled value functions: Specifically, first, in the framework of approximate dynamic programming, the Markov decision process described by Equations (1) to (25) can be decomposed into a series of solvable sub-problems by the Bellman equation; taking the t-th sub-period to be optimized as an example, since the decision variables (i.e., the parameters to be optimized) x of the t-th sub-period to be optimized are known, t , the state space parameter S of the tth sub-period to be optimized t , the state transition parameter W of the tth sub-period to be optimized t And the objective function, thus decomposing equations (1) to (25) to obtain equation (26), where the first value function V t (S t ) is used to describe the interaction between each sub-problem, and its specific expression is as follows: Where, E[·] represents the expectation operation; However, due to the existence of external information, the subproblem defined by Equation (26) may be affected by the curse of dimensionality, so it is necessary to introduce the post-decision state to deal with this difficulty, which can be written as: Where, is the post-decision value function of the (t-Δt)th sub-period to be optimized; is the state space parameter of the tth sub-period to be optimized after the optimization decision is made and before the state transfer parameter is obtained; is the post-decision value function of the tth sub-period to be optimized; is random external information W t The profit under x t is the decision variable for the tth sub-period to be optimized; is the state transition parameter of the nth power operation scenario in the tth sub-period to be optimized; is the state space parameter of the (t-Δt)th sub-period to be optimized.

[0030] In order to deal with the randomness of external information in Equation (27), the Monte Carlo method is used to transform Equation (27) into Equation (28): in, represents the value function obtained by the Monte Carlo method; is the value function obtained by Monte Carlo method; is the nth sampling scene W t n The income under N W Run scenario sets for electricity; Specifically, in formula (28), in the tth sub-period to be optimized, the optimization problem corresponding to the nth power operation scenario in the power operation scenario set is written as: That is, formula (29) is the expression of the decoupled value function. Thus, the decoupling of the initial optimization model is completed, and the optimization model represented by several decoupled value functions is obtained: In a preferred embodiment, the expression of the decoupled value function is specifically: in, is the value function of the (t-Δt)th sub-period to be optimized in the nth power operation scenario after decoupling, is the state space parameter of the (t-Δt)th sub-period to be optimized; is the instantaneous profit function, which represents the profit of the (t-Δt)th sub-period to be optimized in the nth power operation scenario, x t is the parameter to be optimized in the tth sub-period to be optimized, W t n is the state transition parameter of the nth power operation scenario in the tth sub-period to be optimized; is the value function obtained by Monte Carlo method, is the state space parameter of the tth sub-period to be optimized after the optimization decision is made and before the state transfer parameter is updated; T is the set of sub-periods to be optimized, which consists of all sub-periods to be optimized; NW Run scenario sets for electricity.

[0031] S3. Based on the optimization constraints, convert each decoupled value function into a corresponding Lagrangian function, and calculate the gradient of each Lagrangian function to obtain the KKT condition corresponding to each Lagrangian function; For example, since the value function is implicitly unknown, existing studies typically use piecewise linear functions or Benders cuts for approximation. However, these first-order functions cannot well describe the nonlinearity of the problem constructed by the optimization model of this application, resulting in inaccurate solution results. To solve this problem, this embodiment proposes a value function approximation technology based on multidimensional holomorphic embedding: In a preferred embodiment, the optimization constraint further comprises: running inequality constraints; Based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated respectively to obtain the KKT conditions corresponding to each Lagrangian function, including: Introducing a preset slack variable into the running inequality constraint so that the running inequality constraint is converted into an equation form to obtain a second running equality constraint; Determine the lower limit expression of the slack variable according to the preset barrier factor and the preset slack variable value constraint; Based on the first operating equality constraint, the second operating equality constraint, and the slack variable lower limit expression, converting each decoupled value function into a corresponding Lagrangian function; Calculating the gradient of the Lagrangian function to obtain the KKT conditions corresponding to each Lagrangian function; Specifically, a preset slack variable is introduced into the running inequality constraint, so that the running inequality constraint is converted into an equation form: Where, P t for the reason and The vector of components; P t P t The lower limit of is the upper limit of P; Q t for the reason and The vector of components; Q t Q t The lower limit of Q t Upper limit of and is the slack variable of formula (12); and is the slack variable of formula (13); and is the slack variable for running inequality constraints (9), (11) and (14); and is the slack variable for running the inequality constraints (10) and (15).

[0032] Since all the slack variables introduced in equations (30) to (35) have non-negative constraints, they are eliminated by the interior point method. Specifically, the lower limit expression of the slack variables is determined according to the preset barrier factor and the preset slack variable value constraint. It should be noted that the slack variable value constraint constrains all slack variables to be non-negative numbers. (Specifically, based on the first running equation constraint, the second running equation constraint and the slack variable lower limit expression, each decoupled value function is converted into a corresponding Lagrangian function. Thus, the Lagrangian function of equation (29) can be written as: Wherein, l and u are vectors consisting of slack variables of the constraints within the running inequality constraint; λ is a vector consisting of Lagrange multipliers of the constraints within the first running equality constraint; and π are vectors consisting of the Lagrange multipliers of the constraints within the running inequality constraint; is the Lagrange multiplier constraining the energy storage state change; is the first Lagrange multiplier of the active power balance constraint; is the second Lagrange multiplier for the reactive power balance constraint; and is the Lagrange multiplier constraining (29)-(34); N l is the dimension of l; l k is the kth element of l; u k is the kth element of u; μ is the preset barrier factor; is the lower bound expression of the slack variable; Then, the gradient of the Lagrangian function is calculated to obtain the KKT conditions corresponding to each Lagrangian function: Specifically, the expression of the KKT condition is: Where 1 is the unit column vector; U = diag(u); L = diag(l); Π =diag( π ); It should be noted that formula (41) is equivalent to formula (16)-formula (18) and formula (30)-formula (35); Integrating equations (37) and (42), we obtain equation (43): Where,

[0033] S4. Embed the preset complex parameters into each KKT condition, expand the variables in the KKT condition embedded with the complex parameters in the form of a power series, obtain the polynomial equation group corresponding to each KKT condition, and calculate the zero-order coefficient of each polynomial equation group; schematically, since formula (43) only reflects the current state space parameters and the optimal decision a (i.e. ), therefore, it is necessary to apply the multidimensional holomorphic embedding method to obtain its explicit expression; In a preferred embodiment, the complex parameters include a first complex parameter; The embedding of the preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the first complex parameter into each KKT condition, and expanding the variables in the KKT condition embedded in the first complex parameter in a power series form to obtain a first polynomial equation group corresponding to each KKT condition; For each of the first polynomial equation groups, merge like terms of the first polynomial equation group to obtain a first polynomial equation group after merging like terms; extract coefficients of each like term in the first polynomial equation group after merging like terms; construct a coefficient equation group based on preset coefficients and conditions and the coefficients of each like term; solve the coefficients of zero-order like terms in the coefficient equation group to determine the zero-order coefficient of the first polynomial equation group; Obtaining zero-order coefficients of each of the first polynomial equations; For example, let's take a KKT condition as an example: Specifically, the first complex parameter is first embedded into the KKT condition, i.e., Equation (43), to obtain: Where, s=(s1,s2,...,s D ) is the first complex parameter, indicating the proportion of the current state, and D is the dimension of S; Since a(s) is holomorphic, the KKT condition embedded in the first complex parameter is expanded in the form of a power series to obtain the first polynomial equation system: Where ai [n1, n2, ..., n D ]for The coefficient of N a is the dimension of a(·).

[0034] Note that for any value of S, the coefficients of each order on both sides of equation (44) should be equal; Merge like terms of the first polynomial equation group to obtain a first polynomial equation group after merging like terms, extract the coefficients of each like term in the first polynomial equation group after merging like terms, set the sum of the coefficients of like terms in each equation to 0, and obtain a set of coefficient equation groups; assume that the zero-order coefficients on both sides of equation (31) are equal, solve the coefficients of the zero-order like terms in the coefficient equation group, and determine the zero-order coefficients of the first polynomial equation group.

[0035] S5. For each of the polynomial equations, perform a recursive calculation operation on the non-zero-order coefficients of the polynomial equations according to the zero-order coefficients until the solved order of the polynomial equations reaches a preset maximum order, terminate the recursive calculation operation, obtain coefficient solutions of each order of the polynomial equations, and obtain an approximate expression of the value function corresponding to the polynomial equations according to the coefficient solutions; In a preferred embodiment, the recursive calculation operation includes: The order coefficient of the current order to be solved is calculated based on the order coefficient corresponding to the solved order; wherein, when the current order to be solved is first order, the order coefficient corresponding to the solved order is the zero-order coefficient; Determine whether the current order to be solved reaches the maximum order, If not, the current order to be solved is used as the solved order, and the next order after the solved order is used as the current order to be solved for the next recursive calculation operation; If so, terminate the recursive calculation operation; Indicatively, since the zero-order coefficient is the basis for solving the first-order coefficient, and the first-order coefficient is the basis for solving the second-order coefficient, and so on, the solution process of the zero-order coefficient and the recursive calculation operation are explained by taking formula (32) as an example: 1) Derivation of zero-order coefficients: Assuming that the zero-order coefficients on both sides of equation (32) are equal, a set of nonlinear equations can be constructed as follows: By solving these nonlinear equations, the zero-order coefficients are obtained.

[0036] 2) Derivation of the first-order coefficients: Similarly, let the first-order coefficients on both sides of (32) be equal, and we can obtain: Since the zero-order coefficients have been obtained before, Equation (47) becomes a set of linear equations, and the first-order coefficients are calculated based on the linear equations.

[0037] 3) By repeating the above process, coefficients of the second order, third order, and so on can be recursively obtained until the solved order of the polynomial equation system reaches a preset maximum order, terminating the recursive calculation operation, obtaining coefficient solutions of each order of the first polynomial equation system, and obtaining an approximate expression of the value function corresponding to the first polynomial equation system based on the coefficient solutions; 4) According to the coefficient solution, an approximate expression of the value function corresponding to the first polynomial equation group is obtained.

[0038] In the above-mentioned process of embedding the first complex parameter into the KKT condition to calculate the zero-order coefficient of each first polynomial equation group, the calculation time is particularly significant. To solve this problem, this embodiment further accelerates the solution method: in a preferred embodiment, the complex parameter includes a second complex parameter; the optimization constraint includes a first running equation constraint; the first running equation constraint includes an active power balance constraint and a reactive power balance constraint; the KKT condition includes a first Lagrangian multiplier corresponding to the active power balance constraint and a second Lagrangian multiplier corresponding to the reactive power balance constraint; The method of embedding a preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the second complex parameter into each KKT condition, and decoupling the equation with node coupling in the KKT condition embedded in the second complex parameter based on a preset relaxation parameter to obtain a decoupled KKT condition; Expand the variables in each of the decoupled KKT conditions in a power series form to obtain a corresponding second polynomial equation group; Assigning a specific value to the second complex parameter; Assigning the same value to the first Lagrangian multiplier and the second Lagrangian multiplier; Solving each of the second polynomial equation groups to obtain a zero-order coefficient corresponding to each of the second polynomial equation groups according to the second complex parameter assigned a value, the first Lagrange multiplier assigned a value, the second Lagrange multiplier assigned a value, and each of the second polynomial equation groups; Schematically, Equations (16), (17), (37), and (38) contain information about all nodes in the system, which means that all nodes are coupled in these equations. Therefore, it is necessary to introduce a second complex parameter s′=(s,s D+1 ), in order to achieve the purpose of node decoupling in KKT conditions; where s D+1Any value between 0 and 1; Specifically, the second complex parameter s′=(s,s D+1 ) is embedded into each KKT condition, and based on a preset relaxation parameter, the equations with node coupling in the KKT condition embedded in the second complex parameter are decoupled to obtain the decoupled KKT condition, and the following formula is obtained: L(s′) Π (s′)1-(μ+(1-s D+1 )Δμ0)1=0; (53) Where s′=(s,s D+1 ); In order to achieve decoupling when obtaining the 0th order coefficient, we define the relaxation parameter and Δμ0, specifically: Where μ0 is the preset barrier factor when s′=0.

[0039] Specifically, equations (48)-(52) are the relaxed versions of equations (16), (17), (37), (38) and (32), respectively. Equations (53) and (54) are the requirements of the interior point method. Equation (55) is the abstract form of the KKT conditions except (48)-(54) with the second complex parameter embedded. See also Figure 2 , the variables in the decoupled KKT condition are expanded in power series form to obtain the second polynomial equation group, with e i，t (s′) as an example, it is expressed as: Specifically, the second complex parameter s′ is set to 0 to delete the high-order terms and retain only the zero-order terms. Then, Equations (48) and (49) are rewritten as: Obviously, e i，t (0) = 1, f i，t (0) = 0 is the solution of equation (58) - equation (59). Substituting this solution into equations (50) and (51), we obtain: The first Lagrangian multiplier and the second Lagrangian multiplier are given the same value, so as to satisfy equation (62). Since equation (52) relaxes the voltage lower limit, the corresponding Lagrangian multiplier is It can take any value, satisfying formula (63): Where, for The value at all nodes; for The value at all nodes; In this case, equations (60) and (61) can be replaced by fully decoupled equations (62) and (63), and equations (60) and (61) always hold true. Thus, based on equations (53) to (56), (62), and (63), the zero-order coefficients of the second polynomial equation group are finally solved. After the zero-order coefficients are solved, the first-order coefficients are calculated similarly through recursive calculation operations: until the solved order of the second polynomial equation group reaches a preset maximum order, terminating the recursive calculation operation to obtain coefficient solutions of each order of the second polynomial equation group; Obtaining an approximate expression of the value function corresponding to the polynomial equation group according to the coefficient solution; S6. deriving a solution to the optimization model based on the approximate expressions of the value functions, and dispatching the power system based on the solution to the optimization model; Specifically, after obtaining the approximate expressions of each value function, the solution of the optimization model is solved based on a preset solver, and the power system is dispatched according to the solution of the optimization model.

[0040] See Figure 3 , is a power system dispatching device based on multidimensional holomorphic embedding approximate dynamic programming provided by one embodiment of the present invention, comprising: a data acquisition module, an optimization model construction module, a KKT condition determination module, a KKT condition processing module, a value function approximate expression solving module, and a dispatching module; The data acquisition module is used to obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and a dispatch cost coefficient; The optimization model construction module is used to construct optimization constraints based on the acquired data, and to construct an optimization model represented by a plurality of decoupled value functions with the goal of minimizing the total cost of the power system; wherein each of the value functions is used to represent the optimization problem corresponding to a certain power operation scenario in the power operation scenario set within a preset sub-period to be optimized; The KKT condition determination module is used to convert each decoupled value function into a corresponding Lagrangian function based on the optimization constraints, and calculate the gradient of each Lagrangian function to obtain the KKT condition corresponding to each Lagrangian function; The KKT condition processing module is used to embed a preset complex parameter into each KKT condition, expand the variables in the KKT condition embedded with the complex parameter in the form of a power series to obtain a polynomial equation group corresponding to each KKT condition, and calculate the zero-order coefficient of each polynomial equation group; The value function approximate expression solving module is configured to perform a recursive calculation operation on the non-zero-order coefficients of each polynomial equation group according to the zero-order coefficient, until the solved order of the polynomial equation group reaches a preset maximum order, terminate the recursive calculation operation, obtain coefficient solutions of each order of the polynomial equation group, and obtain an approximate expression of the value function corresponding to the polynomial equation group according to the coefficient solutions; The scheduling module is used to solve the solution of the optimization model based on the approximate expressions of each value function, and to schedule the power system based on the solution of the optimization model.

[0041] See also Figure 4 , an embodiment of the present application further provides a terminal device, including: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming as described above.

[0042] The processor is used to control the overall operation of the terminal device to complete all or part of the steps of the above-mentioned power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming. The memory is used to store various types of data to support the operation of the terminal device. These data may include, for example, instructions for any application or method used to operate on the terminal device, as well as application-related data. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0043] In an exemplary embodiment, the terminal device can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the power system scheduling method based on multi-dimensional holomorphic embedded approximate dynamic programming as described in any of the above embodiments, and achieve technical effects consistent with the above methods.

[0044] In another exemplary embodiment, a computer-readable storage medium including a computer program is further provided. When executed by a processor, the computer program implements the steps of the power system scheduling method based on multidimensional holomorphic embedding approximate dynamic programming as described in any of the above embodiments. For example, the computer-readable storage medium may be the aforementioned memory including the computer program. The computer program may be executed by a processor of a terminal device to implement the power system scheduling method based on multidimensional holomorphic embedding approximate dynamic programming as described in any of the above embodiments, and achieve the same technical effects as the above methods.

[0045] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming, characterized in that: include: Obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and dispatch cost coefficients; Based on the acquired data, optimization constraints are established, and with the goal of minimizing the total cost of the power system, an optimization model represented by a plurality of decoupled value functions is constructed; wherein each of the value functions is used to represent the optimization problem corresponding to a power operation scenario in the power operation scenario set within a predetermined sub-period to be optimized; Based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated to obtain the KKT condition corresponding to each Lagrangian function; Embedding a preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded with the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group; For each of the polynomial equations, recursively calculating the non-zero-order coefficients of the polynomial equations according to the zero-order coefficients until the solved order of the polynomial equations reaches a preset maximum order, terminating the recursive calculation operation, obtaining coefficient solutions of each order of the polynomial equations, and obtaining an approximate expression of the value function corresponding to the polynomial equations according to the coefficient solutions; The solution of the optimization model is solved according to the approximate expression of each value function, and the power system is dispatched according to the solution of the optimization model.

2. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 1, characterized in that: The complex parameters include a first complex parameter; The embedding of the preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the first complex parameter into each KKT condition, and expanding the variables in the KKT condition embedded in the first complex parameter in a power series form to obtain a first polynomial equation group corresponding to each KKT condition; For each of the first polynomial equation groups, merge like terms of the first polynomial equation group to obtain a first polynomial equation group after merging like terms; extract coefficients of each like term in the first polynomial equation group after merging like terms; construct a coefficient equation group based on preset coefficients and conditions and the coefficients of each like term; solve the coefficients of zero-order like terms in the coefficient equation group to determine the zero-order coefficient of the first polynomial equation group; The zero-order coefficients of each of the first polynomial equations are obtained.

3. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 1, characterized in that: The complex parameter includes a second complex parameter; the optimization constraint includes a first operating equation constraint; the first operating equation constraint includes an active power balance constraint and a reactive power balance constraint; the KKT condition includes a first Lagrangian multiplier corresponding to the active power balance constraint and a second Lagrangian multiplier corresponding to the reactive power balance constraint; The method of embedding a preset complex parameter into each KKT condition, expanding the variables in the KKT condition embedded in the complex parameter in a power series form to obtain a polynomial equation group corresponding to each KKT condition, and calculating the zero-order coefficient of each polynomial equation group includes: embedding the second complex parameter into each KKT condition, and decoupling the equation with node coupling in the KKT condition embedded in the second complex parameter based on a preset relaxation parameter to obtain a decoupled KKT condition; Expanding the variables in each of the decoupled KKT conditions in a power series form to obtain a corresponding second polynomial equation group; assigning specific values to the second complex parameters; Assigning the same value to the first Lagrangian multiplier and the second Lagrangian multiplier; According to the second complex parameter after being assigned a value, the first Lagrange multiplier after being assigned a value, the second Lagrange multiplier after being assigned a value and each of the second polynomial equation groups, the zero-order coefficient corresponding to each of the second polynomial equation groups is obtained by solving.

4. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 1, characterized in that: The recursive calculation operation includes: The order coefficient of the current order to be solved is calculated based on the order coefficient corresponding to the solved order; wherein, when the current order to be solved is first order, the order coefficient corresponding to the solved order is the zero-order coefficient; Determine whether the current order to be solved reaches the maximum order, If not, the current order to be solved is used as the solved order, and the next order after the solved order is used as the current order to be solved for the next recursive calculation operation; If so, terminate the recursive calculation operation.

5. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 1, characterized in that: Each power operation scenario includes a set of state space parameters and a set of state transition parameters; the parameters to be optimized include the output of the thermal power unit, the charge and discharge power of the energy storage power station in each node, the real and imaginary components of the voltage vector of each node, and the transmission power used for power interaction with the upper power grid; The optimization constraints are constructed based on the acquired data, and with the goal of minimizing the total cost of the power system, an optimization model represented by several decoupled value functions is constructed, including: Based on the acquired data, optimization constraints are established, and an initial optimization model is constructed using the output of the thermal power units, the charge and discharge power of the energy storage power station at each node, the real and imaginary components of the voltage vector at each node, and the transmission power for power interaction with the upper power grid as decision variables, with the goal of minimizing the total cost of the power system; According to a preset approximate dynamic programming algorithm, the initial optimization model is decoupled to obtain the optimization model represented by the plurality of decoupled value functions.

6. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 5, characterized in that: The expression of the decoupled value function is specifically: in, is the value function of the (t-Δt)th sub-period to be optimized in the nth power operation scenario after decoupling, is the state space parameter of the (t-Δt)th sub-period to be optimized; is the instantaneous profit function, which represents the profit of the (t-Δt)th sub-period to be optimized in the nth power operation scenario, x t is the parameter to be optimized in the tth sub-period to be optimized, W t n is the state transition parameter of the nth power operation scenario in the tth sub-period to be optimized; is the value function obtained by Monte Carlo method, is the state space parameter of the tth sub-period to be optimized after the optimization decision is made and before the state transfer parameter is updated; T is the set of sub-periods to be optimized, which consists of all sub-periods to be optimized; N W Run scenario sets for electricity.

7. The power system dispatching method based on multidimensional holomorphic embedding approximate dynamic programming according to claim 3, characterized in that: The optimization constraints further include: running inequality constraints; Based on the optimization constraints, each decoupled value function is converted into a corresponding Lagrangian function, and the gradient of each Lagrangian function is calculated respectively to obtain the KKT conditions corresponding to each Lagrangian function, including: Introducing a preset slack variable into the running inequality constraint so that the running inequality constraint is converted into an equation form to obtain a second running equality constraint; Determine the lower limit expression of the slack variable according to the preset barrier factor and the preset slack variable value constraint; Based on the first operating equality constraint, the second operating equality constraint, and the slack variable lower limit expression, converting each decoupled value function into a corresponding Lagrangian function; The gradient of the Lagrangian function is calculated to obtain the KKT conditions corresponding to each Lagrangian function.

8. A power system dispatching device based on multidimensional holomorphic embedding approximate dynamic programming, characterized in that: include: Data acquisition module, optimization model construction module, KKT condition determination module, KKT condition processing module, value function approximate expression solution module and scheduling module; The data acquisition module is used to obtain a set of power operation scenarios, parameters to be optimized, a set of power equipment, a set of nodes, operating limits of various types of power equipment, grid interaction power limits, inter-node line admittance parameters, and a dispatch cost coefficient; The optimization model construction module is used to construct optimization constraints based on the acquired data, and to construct an optimization model represented by a plurality of decoupled value functions with the goal of minimizing the total cost of the power system; wherein each of the value functions is used to represent the optimization problem corresponding to a certain power operation scenario in the power operation scenario set within a preset sub-period to be optimized; The KKT condition determination module is used to convert each decoupled value function into a corresponding Lagrangian function based on the optimization constraints, and calculate the gradient of each Lagrangian function to obtain the KKT condition corresponding to each Lagrangian function; The KKT condition processing module is used to embed a preset complex parameter into each KKT condition, expand the variables in the KKT condition embedded with the complex parameter in the form of a power series to obtain a polynomial equation group corresponding to each KKT condition, and calculate the zero-order coefficient of each polynomial equation group; The value function approximate expression solving module is configured to perform a recursive calculation operation on the non-zero-order coefficients of each polynomial equation group according to the zero-order coefficient, until the solved order of the polynomial equation group reaches a preset maximum order, terminate the recursive calculation operation, obtain coefficient solutions of each order of the polynomial equation group, and obtain an approximate expression of the value function corresponding to the polynomial equation group according to the coefficient solutions; The scheduling module is used to solve the solution of the optimization model based on the approximate expressions of each value function, and to schedule the power system based on the solution of the optimization model.

9. A terminal device, characterized in that: include: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power system scheduling method based on multi-dimensional holomorphic embedding approximate dynamic programming as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the power system scheduling method based on multidimensional holomorphic embedding approximate dynamic programming as described in any one of claims 1 to 7 is implemented.

Images