Regulation and control method and device of electricity-hydrogen hybrid energy storage system, computer equipment and storage medium

Abstract

The invention relates to a regulation and control method and device for an electricity-hydrogen hybrid energy storage system, computer equipment and a storage medium. The method comprises the steps that distribution network information of the electricity-hydrogen hybrid energy storage system is acquired; the distribution network information comprises source load data, operation parameters and distribution network topology information of the electricity-hydrogen hybrid energy storage system; establishing a regulation and control model of the electricity-hydrogen hybrid energy storage system, and dividing the regulation and control model into a plurality of sub-regulation and control models according to a preset time period; each sub-regulation model comprises a coupling constraint, and the coupling constraint is a constraint between the sub-regulation model and the adjacent sub-regulation model; determining a plurality of target regulation and control strategies of the electricity-hydrogen hybrid energy storage system according to the distribution network information and the plurality of sub-regulation and control models; the sub-regulation models are in one-to-one correspondence with the target regulation strategies. By using the regulation and control method provided by the invention, the rate of determining the target regulation and control strategy can be improved.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a control method, apparatus, computer equipment, and storage medium for an electro-hydrogen hybrid energy storage system. Background Technology

[0002] Against the backdrop of the accelerated transformation of the global energy structure, the integration of a high proportion of renewable energy into the distribution network has become a key path to promote the clean and low-carbon development of the power system. Advances in power electronic conversion technology and intelligent control systems have provided technical support for the coordinated operation of hybrid electric-hydrogen energy storage systems.

[0003] In traditional technologies, power flow calculation analysis is typically used to analyze the control strategies of hybrid electric-hydrogen energy storage systems. However, the solution rate of traditional power flow calculation analysis methods is relatively slow. Summary of the Invention

[0004] Therefore, it is necessary to provide a control method, apparatus, computer equipment, and storage medium for an electric-hydrogen hybrid energy storage system that can improve the rate of determining the control strategy, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for regulating an electro-hydrogen hybrid energy storage system, the method comprising:

[0006] Obtain the distribution network information of the hybrid electric-hydrogen energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the hybrid electric-hydrogen energy storage system.

[0007] A control model for an electric-hydrogen hybrid energy storage system is established, and the control model is divided into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0008] Based on distribution network information and multiple sub-control models, several target control strategies for the electric-hydrogen hybrid energy storage system are determined; each sub-control model corresponds one-to-one with a target control strategy.

[0009] In one embodiment, based on distribution network information and multiple sub-control models, several target control strategies for the electric-hydrogen hybrid energy storage system are determined, including:

[0010] Based on distribution network information, initial control strategies, and multiple sub-control models, multiple intermediate control strategies are determined;

[0011] A global quadratic programming model is established based on multiple intermediate control strategies, and the global quadratic programming model is solved based on multiple intermediate control strategies to obtain the control strategy correction amount.

[0012] Based on the adjustment amount of the control strategy and multiple intermediate control strategies, multiple target control strategies are determined.

[0013] In one embodiment, based on distribution network information, an initial control strategy, and multiple sub-control models, multiple intermediate control strategies are determined, including:

[0014] Based on the augmented Lagrange method, the augmented Lagrange model of multiple sub-regulation models is determined;

[0015] Based on distribution network information, initial control strategies, and multiple augmented Lagrange models, several intermediate control strategies are determined.

[0016] In one embodiment, based on distribution network information, an initial control strategy, and multiple augmented Lagrange models, multiple intermediate control strategies are determined, including:

[0017] Based on the Markov decision process, multiple first state spaces are constructed according to distribution network information, initial control strategies, and multiple augmented Lagrange models;

[0018] Multiple first state spaces are input into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0019] In one embodiment, a global quadratic programming model is established based on multiple intermediate control strategies, including:

[0020] Based on multiple intermediate control strategies, the gradient and second-order information of each sub-control model are determined;

[0021] A global quadratic programming model is established based on multiple intermediate control strategies and the gradients and second-order information of all sub-control models.

[0022] In one embodiment, a global quadratic programming model is solved based on multiple intermediate control strategies to obtain the control strategy correction amount, including:

[0023] Based on the Markov decision process, a second state space is constructed according to multiple intermediate control strategies, the gradients and second-order information of all sub-control models, and the global quadratic programming model.

[0024] The second state space is input into the pre-trained second deep reinforcement learning model to obtain the adjustment amount of the regulation policy.

[0025] In one embodiment, multiple target control strategies are determined based on the control strategy correction amount and multiple intermediate control strategies, including:

[0026] Each intermediate control strategy is modified based on the adjustment amount of the control strategy to obtain multiple modified control strategies;

[0027] When multiple corrective control strategies converge, these multiple corrective control strategies are identified as the target control strategy.

[0028] If multiple modified control strategies fail to converge, the multiple modified control strategies are determined as new initial control strategies, and the process returns to the step of determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models.

[0029] Secondly, one embodiment of this application provides a control device for an electro-hydrogen hybrid energy storage system, the device comprising:

[0030] The acquisition module is used to acquire the distribution network information of the electric-hydrogen hybrid energy storage system. The distribution network information includes the source-load data, operating parameters and distribution network topology information of the electric-hydrogen hybrid energy storage system.

[0031] A module is established to build a control model for an electric-hydrogen hybrid energy storage system. The control model is divided into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints, which are the constraints between the sub-control model and adjacent sub-control models.

[0032] The determination module is used to determine multiple target control strategies for the electric-hydrogen hybrid energy storage system based on distribution network information and multiple sub-control models; each sub-control model corresponds one-to-one with a target control strategy.

[0033] Thirdly, one embodiment of this application provides a computer device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the control method provided in the first aspect.

[0034] Fourthly, one embodiment of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the control method provided in the first aspect above.

[0035] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the control method provided in the first aspect above.

[0036] The aforementioned regulation method, apparatus, computer equipment, and storage medium for the hybrid electric-hydrogen energy storage system acquire the distribution network information of the hybrid electric-hydrogen energy storage system. This distribution network information includes source-load data, operating parameters, and distribution network topology information of the hybrid electric-hydrogen energy storage system. A regulation model for the hybrid electric-hydrogen energy storage system is established, and this model is divided into multiple sub-regulation models according to a preset time period. Each sub-regulation model includes coupling constraints, which are constraints between the sub-regulation model and adjacent sub-regulation models. Based on the distribution network information and the multiple sub-regulation models, multiple target regulation strategies for the hybrid electric-hydrogen energy storage system are determined. Each sub-regulation model corresponds one-to-one with a target regulation strategy. In this embodiment, after establishing the regulation model of the hybrid electric-hydrogen energy storage system, it is first divided into multiple sub-regulation models. Multiple target regulation strategies for the hybrid electric-hydrogen energy storage system are determined by solving each sub-regulation model. This improves the rate at which the target regulation strategies are determined, enabling instantaneous autonomous operation decisions in the hybrid electric-hydrogen balance zone, supporting energy management and optimized operation in the hybrid electric-hydrogen balance zone, and thus improving the practicality of the regulation method for the hybrid electric-hydrogen energy storage system. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the structure of the electro-hydrogen balance region in one embodiment;

[0038] Figure 2 This is a flowchart illustrating the steps of the control method in one embodiment;

[0039] Figure 3 This is a flowchart illustrating the steps of the control method in another embodiment;

[0040] Figure 4 This is a flowchart illustrating the steps of the control method in another embodiment;

[0041] Figure 5 This is a flowchart illustrating the steps of the control method in another embodiment;

[0042] Figure 6 This is a flowchart illustrating the steps of the control method in another embodiment;

[0043] Figure 7 This is a flowchart illustrating the steps of the control method in another embodiment;

[0044] Figure 8 This is a flowchart illustrating the steps of the control method in another embodiment;

[0045] Figure 9 This is a flowchart illustrating the steps of the control method in another embodiment;

[0046] Figure 10 This is a schematic diagram of the control device in one embodiment;

[0047] Figure 11 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0050] First, before introducing the technical solutions of the embodiments disclosed in this application, the background technology or technological evolution on which the embodiments of this application are based will be introduced. Against the backdrop of the current accelerated transformation of the global energy structure, the integration of a high proportion of renewable energy into the distribution network has become a key path to promote the clean and low-carbon development of the power system. The large-scale grid connection of distributed new energy sources such as photovoltaics and wind power has significantly improved the green energy supply capacity of the distribution network. However, the intermittency, volatility, and randomness of their output also pose serious challenges to the safe and stable operation of the power grid. In order to improve the distribution network system's ability to absorb new energy and its operational flexibility, battery energy storage and hydrogen energy storage technologies are gradually being applied to the distribution network. Battery energy storage has a rapid response and high regulation accuracy, making it suitable for smoothing power fluctuations and achieving short-term load balancing; hydrogen energy storage, with its advantages of large-scale and long-term storage, can achieve energy transfer and cross-seasonal regulation across multiple time scales. Advances in power electronic conversion technology and intelligent control systems have provided technical support for the coordinated operation of hybrid electric-hydrogen energy storage systems. By creating a novel equilibrium region through the fusion of electricity and hydrogen, and managing and optimizing its energy operation, the reliability, economy, and flexibility of distribution network operation can be enhanced, promoting efficient energy utilization and low-carbon transformation of the system under conditions of high proportion of new energy sources. Traditional technologies typically employ power flow calculation analysis to analyze the control strategy of hybrid electric-hydrogen energy storage systems. However, power flow calculation in power systems is a nonlinear programming problem. When the number of distribution network nodes is too large and the control time is too long, the control model corresponding to the hybrid electric-hydrogen energy storage system will experience dimensionality explosion, resulting in a surge in solution complexity and a slow solution rate. Furthermore, for hybrid electric-hydrogen energy storage systems, the strong randomness of distributed resources, the diverse operating characteristics of each hybrid electric-hydrogen device, and the varied and complex operating conditions of the distribution network make capacity management of hybrid electric-hydrogen energy storage systems difficult, requiring high speed for autonomous operation strategies. Therefore, this application provides a control method for hybrid electric-hydrogen energy storage systems that can improve the solution rate.

[0051] The regulation method for the hybrid energy storage system provided in this application can be applied to distribution network systems connected to a large number of distributed hydrogen energy resources, i.e., hydrogen-electricity balance zones. Figure 1As shown. The hydrogen-electricity balance zone integrates distributed renewable energy generation, such as wind turbines (WT) and photovoltaics (PV), capturing wind and solar energy and converting it into green electricity, which is then supplied to the power grid after conversion by inverters. Battery energy storage (BES) can be charged when renewable energy is abundant, storing surplus energy in the batteries; and discharged when renewable energy is insufficient to meet the electricity load, thus mitigating short-term imbalances between source and load. The electrolyzer (EZ) can drive the internal fuel cell stack to electrolyze water to produce hydrogen when renewable energy is abundant. The produced hydrogen can supply the hydrogen load demand of chemical plants, hydrogen fuel cell vehicles, etc., and can also be stored long-term in a hydrogen storage tank (HST). If the EZ cannot produce enough hydrogen or cannot produce hydrogen due to renewable energy scarcity, hydrogen can be released from the HST to meet the hydrogen load demand. The released hydrogen can also support fuel cell (FC) power generation to supply power to the grid, and work with BES to achieve long-term autonomous operation of the balance zone. Figure 1 The numbers in the diagram represent multiple nodes in the power distribution network. Rectangles represent wind turbines, triangles represent energy storage, ellipses represent photovoltaics, crosses represent hydrogen storage tanks, polygons represent electrolyzers, and circles represent fuel cells. The power distribution network system includes computer equipment, which may include, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices.

[0052] In one embodiment, such as Figure 2 As shown, a method for controlling an electro-hydrogen hybrid energy storage system is provided, and the method is illustrated using a computer device as an example. In this embodiment, the method includes the following steps:

[0053] Step 200: Obtain the distribution network information of the electric-hydrogen hybrid energy storage system; the distribution network information includes the source-load data, operating parameters and distribution network topology information of the electric-hydrogen hybrid energy storage system.

[0054] The source-load data of an electric-hydrogen hybrid energy storage system refers to the power supply-side and load-side data within the system. This data can include active power from wind turbines, active power from photovoltaic power generation, active load, reactive power from wind turbines, reactive power from photovoltaic power generation, reactive load, and hydrogen load demand. The operating parameters of the electric-hydrogen hybrid energy storage system include the rated capacity of the hydrogen storage tank and the rated capacity of the electrical storage. The distribution network topology information of the electric-hydrogen hybrid energy storage system includes the real and imaginary parts of the node admittance matrix of the corresponding distribution network. This distribution network information can be pre-stored in the memory of a computer device, and can be directly retrieved from memory when adjustments to the system are needed.

[0055] Step 210: Establish a control model for the electric-hydrogen hybrid energy storage system, and divide the control model into multiple sub-control models according to a preset time period; each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0056] The comprehensive operating cost of the electro-hydrogen balance zone within the total control time T. To minimize the impact of power consumption on the energy storage system and considering safety constraints related to distribution networks and equipment, a control model for an electric-hydrogen hybrid energy storage system is established. This model includes an objective function and multiple constraints. These constraints include distribution network power flow constraints, electric energy storage operation constraints, hydrogen energy storage operation constraints, and hydrogen energy storage balance constraints.

[0057] Distribution network power flow constraints include:

[0058] (1)

[0059] (2)

[0060] (3)

[0061] (4)

[0062] (5)

[0063] (6)

[0064] (7)

[0065] (8)

[0066] Among them, formulas (1) and (2) are the active power balance constraint and reactive power balance constraint injected at the node, respectively; formulas (3) and (4) are the power flow equations in polar coordinate form, respectively; formulas (5) and (6) are the voltage amplitude constraint and phase angle range constraint at the node, respectively; and formulas (7) and (8) are the active load reduction range constraint and reactive load reduction range constraint at the node, respectively. and Let represent the injected active power and reactive power at node i at time t, respectively. , , , These are the 0-1 parameters for whether node i is connected to wind turbines, photovoltaics, hydrogen energy storage, and electric energy storage, respectively. , , , , , , , These represent the active power generated by the wind turbine, the active power generated by the photovoltaic power generation, the power generated by the fuel cell, the electrolysis power of the electrolyzer, the discharge power of the energy storage, the charging power of the energy storage, the active load, and the reduction in active load at node i at time t. , , , These represent the reactive power generated by the wind turbine, the reactive power generated by the photovoltaic power generation, the reactive load, and the reactive load reduction at node i at time t, respectively. and These are the real and imaginary parts of the element in the i-th row and i-th column of the nodal admittance matrix, respectively. and Let be the voltage magnitude and phase angle of node i at time t, respectively; Let i be the set of nodes connected to node i. and These are the upper and lower limits of the voltage amplitude at node i, respectively; and These are the upper and lower limits of the voltage phase angle at node i, respectively; and These represent the maximum reduction in active load and the maximum reduction in reactive load at node i at time t, respectively.

[0067] Energy storage operation constraints include:

[0068] (9)

[0069] (10)

[0070] (11)

[0071] (12)

[0072] Among them, formulas (9) and (10) are the range constraints of the charging and discharging power of the energy storage, formula (11) is the state of charge (SOC) constraint of the energy storage, and formula (12) is the SOC range constraint of the energy storage. The maximum charging and discharging power of the energy storage on node i; Let SOC be the energy stored at node i at time t; , , , These are the self-discharge rate, rated capacity, charging efficiency, and discharging efficiency of the energy storage at node i, respectively. and These represent the upper and lower limits of the State of Charge (SOC) for energy storage at node i, respectively.

[0073] Hydrogen storage operation constraints include:

[0074] (13)

[0075] (14)

[0076] (15)

[0077] (16)

[0078] (17)

[0079] (18)

[0080] (19)

[0081] (20)

[0082] Among them, formulas (13) and (14) are the range constraints of fuel cell power generation and electrolysis power of electrolyzer, respectively; formulas (15) and (16) are the hydrogen-electric conversion relationship constraints of fuel cell and electrolyzer, respectively; formulas (17) and (18) are the range constraints of hydrogen charging and discharging rate of hydrogen storage tank, respectively; formula (19) is the state of energy (SOE) constraint of hydrogen storage tank; and formula (20) is the SOE range constraint of hydrogen storage tank. and These are the rated power generation of the fuel cell and the rated electrolysis power of the electrolyzer at node i, respectively. and These represent the hydrogen consumption of the fuel cell and the hydrogen production of the electrolyzer at node i at time t, respectively. and These represent the fuel cell power generation efficiency and the electrolyzer electrolysis efficiency at node i, respectively; Q H This refers to the calorific value of hydrogen. and These represent the hydrogen release rate and the amount of hydrogen added to the hydrogen storage tank at node i at time t, respectively. The rated hydrogen charge / discharge capacity of the hydrogen storage tank at node i; Let SOE be the hydrogen storage tank at node i at time t; The rated capacity of the hydrogen storage tank at node i; and These are the upper and lower limits of SOE for the hydrogen storage tank at node i, respectively.

[0083] Hydrogen energy storage balance constraints include:

[0084] (twenty one)

[0085] (twenty two)

[0086] Among them, formula (21) is the hydrogen energy storage conservation constraint, and formula (22) is the hydrogen load reduction range constraint. , , These represent the hydrogen load demand, hydrogen load reduction, and maximum hydrogen load reduction at node i at time t, respectively.

[0087] The objective function in the regulation model of the electric-hydrogen hybrid energy storage system can be expressed as:

[0088] (twenty three)

[0089] in, For network loss costs, To reduce penalties for electricity load, This is the penalty coefficient for hydrogen load reduction.

[0090] Optionally, the objective function may also include the cost of energy storage losses. Electrolytic cell loss cost and fuel cell power generation costs In other words, the objective function can be expressed as:

[0091] (twenty four)

[0092] The energy storage loss cost can be expressed as:

[0093] (25)

[0094] The cost of electrolytic cell losses can be expressed as:

[0095] (26)

[0096] The cost of fuel cell power generation can be expressed as:

[0097] (27)

[0098] Network loss cost can be expressed as:

[0099] (28)

[0100] The penalty cost for reducing electricity load can be expressed as:

[0101] (29)

[0102] The penalty cost of reducing hydrogen load can be expressed as:

[0103] (30)

[0104] The energy storage loss coefficient is... For the set of distribution network nodes, This is the loss coefficient of the electrolytic cell. For fuel cell power generation coefficient, This is the network loss coefficient. The load reduction penalty factor. This is the penalty coefficient for hydrogen load reduction.

[0105] After establishing the control model of the hybrid electric-hydrogen energy storage system, the computer equipment operates according to a preset time period. The regulation model is divided into Multiple sub-control models. The preset time period can be pre-stored in the computer's memory. Assuming the total control period is 12 months, the preset time period can be one month. That is, the control model is divided into 12 sub-control models, with one sub-control model corresponding to one month.

[0106] After dividing the regulation model into multiple sub-regulation models, each sub-regulation model is highly independent and interconnected only through energy storage devices (the energy conservation constraints of electrical energy storage and hydrogen energy storage, i.e., coupling constraints). That is, in the first... In the individual control model, the SOC and SOE of the energy storage device at the end time are compared with those at the second time. The SOC and SOE of the +1 sub-control model are equal at the initial time, as shown in the following formula:

[0107] (31)

[0108] (32)

[0109] The equivalent form of the regulation model can be expressed as:

[0110] (33)

[0111] (34)

[0112] (35)

[0113] Formula (33) is an equivalent form of formula (23) or (24), that is, the sum of the objective functions of multiple sub-control models. Formula (34) is the coupling constraint of each sub-control model, and formula (35) is the constraint condition followed by each sub-control model. The control strategy (decision variable) of the control model. For the first The objective function of the individual regulation model For the first Regulation strategies of individual regulation models For the first The mapping matrix of the sub-control model is responsible for selecting the control strategies participating in the coupling constraints; b is the coupling constraint constant vector; These are the Lagrange multipliers corresponding to the coupling constraints; For the first The constraints followed by the individual regulation model include the multiple constraints provided above; For the first The Lagrange multipliers of the constraints followed by the individual control model.

[0114] Step 220: Based on the distribution network information and multiple sub-control models, determine multiple target control strategies for the electric-hydrogen hybrid energy storage system; each sub-control model corresponds one-to-one with a target control strategy.

[0115] After obtaining multiple sub-control models, the computer equipment solves and analyzes these models based on distribution network information to determine multiple target control strategies for the hybrid electric-hydrogen energy storage system, i.e., the decision variables in the control models. Each sub-control model corresponds to one target control strategy. This embodiment does not limit the specific method for determining multiple target control strategies for the hybrid electric-hydrogen energy storage system based on distribution network information and multiple sub-control models, as long as the function can be achieved.

[0116] The regulation method for an electric-hydrogen hybrid energy storage system provided in this application involves acquiring the distribution network information of the electric-hydrogen hybrid energy storage system. This information includes source-load data and operating parameter load-distribution network topology information. A regulation model for the electric-hydrogen hybrid energy storage system is established, and this model is divided into multiple sub-regulation models according to a preset time period. Each sub-regulation model includes coupling constraints, which are constraints between the sub-regulation model and adjacent sub-regulation models. Based on the distribution network information and the multiple sub-regulation models, multiple target regulation strategies for the electric-hydrogen hybrid energy storage system are determined. Each sub-regulation model corresponds one-to-one with a target regulation strategy. In this embodiment, after establishing the regulation model for the electric-hydrogen hybrid energy storage system, it is first divided into multiple sub-regulation models. Multiple target regulation strategies for the electric-hydrogen hybrid energy storage system are determined by solving each sub-regulation model. This improves the rate at which the target regulation strategies are determined, enabling instantaneous autonomous operation decisions in the electric-hydrogen balance zone, supporting energy management and optimized operation in the electric-hydrogen balance zone, and thus improving the practicality of the regulation method for the electric-hydrogen hybrid energy storage system.

[0117] In an optional embodiment, the computer device may employ the Augmented Lagrangian Alternating Direction Inexact Newton (ALADIN) method to perform distributed solution on the established control model of the electric-hydrogen hybrid energy storage system in order to determine the target control strategy of the electric-hydrogen hybrid energy storage system.

[0118] In one embodiment, such as Figure 3 As shown, this involves an implementation method for determining multiple target control strategies for an electric-hydrogen hybrid energy storage system based on distribution network information and multiple sub-control models. The steps of this implementation method include:

[0119] Step 300: Based on the distribution network information, the initial control strategy, and multiple sub-control models, determine multiple intermediate control strategies.

[0120] The initial control strategy can be pre-set by the user or it can be the control strategy obtained from the previous control method. For each sub-control model, the computer equipment determines the intermediate control strategy based on the distribution network information, the initial control strategy, and the sub-control model, thus obtaining multiple intermediate control strategies.

[0121] Step 310: Establish a global quadratic programming model based on multiple intermediate control strategies, and solve the global quadratic programming model based on multiple intermediate control strategies to obtain the control strategy correction amount.

[0122] After obtaining multiple intermediate control strategies, the computer equipment establishes a global quadratic programming model based on these strategies. The essence of global quadratic programming is to find the global optimum of the quadratic programming model under constraints, that is, to find the optimal control strategy correction amount for the intermediate control strategies corresponding to the sub-control models. In other words, after obtaining the global quadratic programming model, the computer equipment can solve the global quadratic programming model based on multiple intermediate control strategies to obtain the control strategy correction amount.

[0123] Step 320: Determine multiple target control strategies based on the adjustment amount of the control strategy and multiple intermediate control strategies.

[0124] After obtaining the adjustment amount of the control strategy, the computer equipment can obtain multiple target control strategies based on each intermediate control strategy according to the adjustment amount of the control strategy.

[0125] In this embodiment, multiple intermediate control strategies are determined based on distribution network information, initial control strategies, and multiple sub-control models. A global quadratic programming model is established based on these intermediate control strategies, and the model is solved to obtain control strategy correction amounts. Based on the control strategy correction amounts and the multiple intermediate control strategies, multiple target control strategies are determined. By solving the global quadratic programming model to determine the control strategy correction amounts and using these correction amounts to adjust the multiple intermediate control strategies, the problem of the determined target control strategies falling into local optima can be avoided, thereby improving the accuracy of the determined target control strategies.

[0126] In one embodiment, such as Figure 4 As shown, this involves an implementation method for determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models. The steps of this implementation method include:

[0127] Step 400: Determine the augmented Lagrange model for multiple sub-regulation models using the augmented Lagrange method.

[0128] For each sub-regulation model, the augmented Lagrangian model for each sub-regulation model is determined using the augmented Lagrangian method. The augmented Lagrangian model for each sub-regulation model includes the augmented Lagrangian model function and the constraints it follows.

[0129] In an optional embodiment, in the first Adding a penalty term to the objective function in the individual control model yields the augmented Lagrange model function.

[0130] The augmented Lagrange model function can be expressed as:

[0131] (36)

[0132] The constraints that must be followed can be expressed as:

[0133] (37)

[0134] in, , , These represent the Lagrange multipliers, penalty coefficients, and the number of elements constrained during the k-th iteration, respectively. Reference values ​​for intermediate control strategies in individual control models; The intermediate control strategy weight matrix is ​​a positive definite diagonal matrix.

[0135] Step 410: Based on the distribution network information, the initial control strategy, and multiple augmented Lagrange models, determine multiple intermediate control strategies.

[0136] After obtaining the augmented Lagrange model for each sub-control model, the computer equipment determines the intermediate control strategy for each augmented Lagrange model based on the distribution network information, the initial control strategy, and the augmented Lagrange model, thereby obtaining multiple intermediate control strategies.

[0137] In an optional embodiment, as can be seen from formulas (36) and (37), when it is necessary to solve for the intermediate control strategy, it is necessary to know the result obtained in the k-th iteration. , , Initial values ​​can be set during the first iteration. , , By solving this problem, we can obtain the result for the (k+1)th iteration. intermediate control strategy and the Lagrange multipliers corresponding to the inequality constraints. .

[0138] The penalty term of the augmented Lagrange model function, after expansion, can be obtained as formula (38). The penalty term can be obtained by setting corresponding weight coefficients for the intermediate control strategy, i.e., the weight matrix. The intermediate control strategies are modified with a focus to guide the solution direction of the control model. Specifically, this application focuses on intermediate control strategies that participate in coupling constraints and are close to the reference value, while other intermediate control strategies satisfy the operational constraints. The intermediate control strategies that participate in coupling constraints are assigned a weight coefficient of 1. At the same time, to ensure the positive definiteness of the weight matrix, the other intermediate control strategies are assigned a very small weight coefficient, such as 1E-6. The weight matrix can be expressed as formula (39).

[0139] (38)

[0140] (39)

[0141] In this embodiment, augmented Lagrange models of multiple sub-control models are determined based on the augmented Lagrange method; multiple intermediate control strategies are determined based on distribution network information, initial control strategies, and multiple augmented Lagrange models. This method for determining intermediate control strategies is quick and easy to implement.

[0142] In one embodiment, such as Figure 5 As shown, this involves an implementation method for determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple augmented Lagrange models. The steps of this implementation method include:

[0143] Step 500: Based on the Markov decision process, construct multiple first state spaces according to the distribution network information, the initial control strategy, and multiple augmented Lagrange models.

[0144] The augmented Lagrange model is a sequential decision model, meaning that the state of the system at the next time step is only related to the state of the system at the previous time step and the control strategy. It conforms to the Markov Decision Process (MDP). An MDP can be composed of a quadruple (S, A, p, r), where S is the state space, A is the action space, p is the state transition probability, and r is the reward obtained by performing the action.

[0145] For each augmented Lagrange model, based on MDP, the computer equipment can construct the first state space of the augmented Lagrange model according to the distribution network information and the initial control strategy, thus obtaining multiple first state spaces.

[0146] In an optional embodiment, the first state space of the augmented Lagrange model includes source-load data, distribution network topology information, the state (operating parameters) of energy storage devices, Lagrange multipliers, penalty coefficients, and intermediate control strategy reference values, which can be represented as:

[0147] (40)

[0148] The first action space includes the output variables of each device, the load reduction amount, and the distribution network operation variables, which can be expressed as:

[0149] (41)

[0150] The second reward function is the opposite of formula (36), and can be expressed as:

[0151] (42)

[0152] Step 510: Input multiple first state spaces into the pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0153] The first deep reinforcement learning model can be pre-trained and stored in the memory of a computer device. The computer device can input multiple first state spaces into the first deep reinforcement learning model to obtain multiple intermediate control policies.

[0154] In an optional embodiment, the method for training the first deep reinforcement learning model includes: training the first deep reinforcement learning model with the objective of maximizing the accumulated expected reward while maximizing the policy entropy, which can be expressed as:

[0155] (43)

[0156] (44)

[0157] in, The optimal strategy for training deep reinforcement learning algorithms (Soft-Actor-Critic, SAC) is... The policy trained for SAC For the expectation, In strategy Below, the distribution of state and action access, The discount factor at the k-th iteration. For temperature parameters, For strategy In state The entropy under the condition of [the target value]. To achieve the above goal, an Actor (policy network)-Critic (value network) framework can be adopted, utilizing neural networks to [process] the soft state-value function. and soft action value function Approximation is performed. Wherein... The parameters are , defined as the expected soft reward obtainable by executing the current strategy, can be updated using formula (45). The purpose of this update process is to make To more accurately reflect the true value of the current strategy.

[0158] (45)

[0159] Where D is the experience replay buffer, used to store the experience tuples generated during training ( , , , ).

[0160] The parameters are It can be used to evaluate the state Next action The quality of the function can be updated using the soft Bellman equation through the following formula (46).

[0161] (46)

[0162] (47)

[0163] in, In the state Next action Then, the target value that the soft action function should achieve.

[0164] During the training process, in order to make the strategy Improve the strategy in the direction of obtaining higher entropy regularization rewards. It can be updated according to formula (48).

[0165] (48)

[0166] in, The KL divergence between the current strategy and the better strategy. This is the normalized partition function.

[0167] When the trained policy can consistently and stably complete its task, the policy is considered to have converged. Formula (49) defines the moving average success rate in the most recent Y training iterations, and Formula (50) is the convergence judgment condition of the model. That is, if the moving average success rate in the most recent Y training iterations exceeds the preset threshold and the success rate variance is less than the preset threshold, then the policy is considered to have converged.

[0168] (49)

[0169] (50)

[0170] in, Let be the moving average success rate at the k-th iteration. Let be the indicator function for the strategy completing the task in the y-th iteration, with 1 for success and 0 for failure. Both v and are pre-defined thresholds, and Var represents the variance.

[0171] In this embodiment, based on the Markov decision process, multiple first state spaces are constructed according to distribution network information, initial control strategies, and multiple augmented Lagrange models. The multiple first state spaces are input into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies. In this way, by combining the Markov decision process and the pre-trained first deep reinforcement learning model to determine multiple intermediate control strategies, the processing speed can be improved, thereby improving the practicality of the control method for the electric-hydrogen hybrid energy storage system.

[0172] In one embodiment, such as Figure 6 As shown, this involves an implementation method for establishing a global quadratic programming model based on multiple intermediate control strategies. The steps of this implementation method include:

[0173] Step 600: Determine the gradient and second-order information of each sub-control model based on multiple intermediate control strategies.

[0174] After obtaining the intermediate control strategies for each sub-control model, the computer equipment determines the gradient and second-order information of the corresponding sub-control model for each intermediate control strategy. The gradient of the sub-control model includes the gradient vector and the Jacobian matrix of the constraints. The gradient vector points to the direction of the fastest growth of the corresponding objective sub-function, and each row in the Jacobian matrix represents the gradient of a constraint. Based on multiple intermediate control strategies, the Lagrangian functions of the corresponding multiple sub-control models are determined. The Hessian matrix is ​​then determined based on the Lagrangian functions of the sub-control models. The Hessian matrix includes the second-order partial derivatives (second-order information) of the Lagrangian functions of the sub-control models, describing the curvature of the Lagrangian functions, i.e., the rate of gradient change.

[0175] No. The gradient vector of the objective function in the individual regulation model It can be represented as:

[0176] (51)

[0177] No. Jacobian matrix of constraints in the individual control model It can be represented as:

[0178] (52)

[0179] No. Lagrangian function of the individual regulation model It can be represented as:

[0180] (53)

[0181] The Hessian matrix corresponding to the Lagrange function It can be represented as:

[0182] (54)

[0183] Step 610: Based on multiple intermediate control strategies and the gradients and second-order information of all sub-control models, establish a global quadratic programming model.

[0184] After obtaining the gradients and second-order information of all sub-control models, the computer equipment establishes a global quadratic programming model based on multiple intermediate control strategies and the gradients and second-order information of all sub-control models.

[0185] In an optional embodiment, the established global quadratic programming model includes: a sub-regulation model in The second-order Taylor expansion formula (55) at the point of intersection, with the constant term removed. The augmented Lagrangian function (56) and the linearized coupling constraints (considering the coupling constraints modified by the intermediate control strategy) can be expressed as formula (57). In order to simplify the global quadratic programming model, in the Jacobian matrix of formula (52), for constraints that are strictly less than 0, i.e. inactive constraints, all their corresponding rows are set to zero, and for constraints that are strictly equal to 0, i.e. active constraints, their corresponding rows remain unchanged. This makes it possible to only consider the modification of the variables of active constraints in the global quadratic programming model, i.e., active constraints are regarded as equality constraints. The first-order Taylor expansion of equality constraints can be expressed as formula (58). Formula (58) removes the constant term. This is expressed as formula (59). The following formula (60) represents a variable that has reached its own boundary value, and its correction amount is 0 to prevent exceeding the limit.

[0186] (55)

[0187] (56)

[0188] (57)

[0189] (58)

[0190] (59)

[0191] (60)

[0192] Where s is a slack variable, allowing violations of the linearization coupling constraint. If the coupling constraint is enforced, i.e., s=0, then There may be no solution. Let be the penalty coefficient for the k-th iteration. For the Lagrange multipliers corresponding to the linearized coupling constraints, and They are respectively The first vector The upper and lower limits of each variable.

[0193] In this embodiment, the gradient and second-order information of each sub-control model are determined based on multiple intermediate control strategies. A global quadratic programming model is established based on each intermediate control strategy, gradient, and second-order information. This method of establishing a global quadratic programming model is quick and easy to implement. Solving the established global quadratic programming model can avoid the problem of multiple target control strategies being trapped in local optima, thereby improving the accuracy of the determined multiple target control strategies.

[0194] In one embodiment, such as Figure 7As shown, this involves solving a global quadratic programming model based on multiple intermediate control strategies to obtain the control strategy correction amount. The steps of this implementation method include:

[0195] Step 700: Based on the Markov decision process, construct the second state space according to multiple intermediate control strategies, the gradients and second-order information of all sub-control models, and the global quadratic programming model.

[0196] The global quadratic programming model is a sequential decision problem, conforming to a Markov decision process. Based on all intermediate control strategies and the gradients and second-order information of the sub-control models corresponding to all intermediate control strategies, the second state space of the global quadratic programming model can be constructed.

[0197] In an optional embodiment, the second state space of the global quadratic programming model includes the Hessian matrix, gradient vector, Lagrange multipliers, penalty coefficient, intermediate control strategy, coupling constraint constant vector, and Jacobian matrix, which can be represented as:

[0198] (61)

[0199] The second action space of the global quadratic programming model, including the adjustment of the control strategy and the slack variables, can be expressed as:

[0200] (62)

[0201] The reward function of the global quadratic programming model is the negative of formula (56), which can be expressed as:

[0202] (63)

[0203] Step 710: Input the second state space into the pre-trained second deep reinforcement learning model to obtain the adjustment policy correction amount.

[0204] The second deep reinforcement learning model can be pre-trained and stored in the memory of a computer device. After obtaining the second state space, the computer device inputs this second state space into the pre-trained second deep reinforcement learning model to obtain the policy adjustment amount. The training process for the second deep reinforcement learning model can refer to the training process for the first deep reinforcement learning model described above.

[0205] In this embodiment, based on the Markov decision process, a second state space is constructed according to the gradient and second-order information of each intermediate control strategy, each sub-control model, and the global quadratic programming model. The second state space is then input into a pre-trained second deep reinforcement learning model to obtain the control strategy correction amount. By combining the Markov decision process and the pre-trained second deep reinforcement learning model to determine the control strategy correction amount, the processing speed can be improved, thereby enhancing the practicality of the control method for the electric-hydrogen hybrid energy storage system.

[0206] In one embodiment, such as Figure 8 As shown, this involves an implementation method for determining a target control strategy based on the control strategy correction amount and multiple intermediate control strategies. The steps of this implementation method include:

[0207] Step 800: Modify each intermediate control strategy according to the control strategy correction amount to obtain multiple modified control strategies.

[0208] For each intermediate control strategy, the computer device uses the control strategy correction amount to modify the intermediate control strategy, thereby obtaining the corresponding modified control strategy, and thus obtaining multiple modified control strategies.

[0209] Specifically, the formula for modifying the intermediate control strategy can be expressed as:

[0210] (64)

[0211] In an optional embodiment, the Lagrange multipliers and penalty coefficients are modified simultaneously with the modification of the intermediate control strategy using the control strategy modification amount, so that they can be used in subsequent iterations.

[0212] The formula for correcting the Lagrange multipliers can be expressed as:

[0213] (65)

[0214] The formula for correcting the penalty coefficient can be expressed as:

[0215] (66)

[0216] in, , , , These are the reference value of the intermediate control strategy used in the (k+1)th iteration, the Lagrange multiplier, and the penalty coefficient, respectively. , , Let the step size be [the value of the step size]. If the full step size is acceptable, then [the value of the step size] can be [the value of the step size]. = = =1; and A fixed coefficient between 0 and 1; and It is a fixed limit.

[0217] In an optional embodiment, as shown in Equation (54), for large programming optimization problems, the computational cost of accurately calculating the Hessian matrix is ​​enormous. A quasi-Newton method (Broyden-Fletcher-Goldfarb-Shanno, BFGS) can be used, eliminating the need to calculate the second derivative and constructing an approximate Hessian matrix using the previous first-order gradient information. The specific calculation method includes: calculating the change in the control strategy between two adjacent iterations, as shown in Equation (67); calculating the gradient change between two adjacent iterations, as shown in Equations (68) and (69), where Equation (69) is the negative gradient of the augmented Lagrange term; and finally, calculating the iterative Hessian approximate matrix, as shown in Equation (70).

[0218] (67)

[0219] (68)

[0220] (69)

[0221] (70)

[0222] in, and The first The change in control strategy and gradient between the (k+1)th iteration and the kth iteration in the sub-control model. The damping coefficient is... For the first The Lagrange gradient of the individual control model at the k-th iteration.

[0223] Step 810: Determine whether multiple corrective control strategies have converged;

[0224] Step 820: When multiple corrective control strategies converge, the corrective control strategy is determined as multiple target control strategies.

[0225] Step 830: If multiple modified control strategies fail to converge, determine the multiple modified control strategies as new initial control strategies, and return to the step of determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models.

[0226] After receiving multiple correction and control strategies, the computer device will determine whether the multiple correction and control strategies have converged. Specifically, this can be determined using the following formulas (71) and (72).

[0227] (71)

[0228] (72)

[0229] in, For convergence accuracy.

[0230] If the computer equipment determines that multiple corrective control strategies have converged, then the multiple corrective control strategies are directly used as multiple target control strategies, so that the electric-hydrogen hybrid energy storage system can send control commands to the corresponding equipment in the distribution network based on the multiple target control strategies.

[0231] If the computer determines that multiple corrective control strategies have not converged, it will use these multiple corrective control strategies as new initial control strategies and return to steps 300-310.

[0232] In this embodiment, each intermediate control strategy is modified according to the correction amount of the control strategy to obtain multiple modified control strategies. If the multiple modified control strategies converge, they are determined as the target control strategy. If the multiple modified control strategies do not converge, they are determined as new initial control strategies, and the above iterative process is repeated. This can ensure superlinear convergence speed while determining an accurate target control strategy, thereby enabling autonomous instantaneous decision-making in the electro-hydrogen equilibrium region, which is helpful for the capacity management and optimized operation of the electro-hydrogen equilibrium region.

[0233] Please see Figure 9 One embodiment of this application provides a control method for an electro-hydrogen hybrid energy storage system, the method comprising the following steps:

[0234] Step 900: Obtain the distribution network information of the electric-hydrogen hybrid energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the electric-hydrogen hybrid energy storage system.

[0235] Step 910: Establish a control model for the electric-hydrogen hybrid energy storage system. Divide the control model into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints. The coupling constraints are the constraints between the sub-control model and the adjacent sub-control models.

[0236] Step 920: Determine the augmented Lagrange model for multiple sub-regulation models using the augmented Lagrange method;

[0237] Step 930: Based on the Markov decision process, construct multiple first state spaces according to the distribution network information, the initial control strategy, and multiple augmented Lagrange models;

[0238] Step 940: Input multiple first state spaces into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies;

[0239] Step 950: Determine the gradient and second-order information of each sub-control model based on multiple intermediate control strategies;

[0240] Step 960: Based on multiple intermediate control strategies and the gradient and second-order information of all sub-control models, establish a global quadratic programming model.

[0241] Step 970: Based on the Markov decision process, construct the second state space according to multiple intermediate control strategies, the gradients and second-order information of all sub-control models, and the global quadratic programming model.

[0242] Step 980: Input the second state space into the pre-trained second deep reinforcement learning model to obtain the adjustment amount of the regulation policy;

[0243] Step 990: Adjust each intermediate control strategy according to the adjustment amount of the control strategy to obtain multiple adjusted control strategies;

[0244] Step 991: Determine whether multiple corrective control strategies have converged;

[0245] Step 992: When multiple corrective control strategies converge, determine the multiple corrective control strategies as the target control strategy.

[0246] Step 993: If multiple corrective control strategies fail to converge, determine the multiple corrective control strategies as the new initial control strategies, and return to execute steps 930-992.

[0247] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0248] Based on the same inventive concept, this application also provides a control device for an electric-hydrogen hybrid energy storage system for implementing the control method of the electric-hydrogen hybrid energy storage system described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of the control device embodiments of the electric-hydrogen hybrid energy storage system provided below can be found in the limitations of the control method of the electric-hydrogen hybrid energy storage system above, and will not be repeated here.

[0249] In one exemplary embodiment, such as Figure 10 As shown, a control device for an electro-hydrogen hybrid energy storage system is provided, comprising: an acquisition module 11, an establishment module 12, and a determination module 13, wherein:

[0250] The acquisition module 11 is used to acquire the distribution network information of the electric-hydrogen hybrid energy storage system. The distribution network information includes the source-load data, operating parameters and distribution network topology information of the electric-hydrogen hybrid energy storage system.

[0251] Module 12 is used to establish a control model for the electric-hydrogen hybrid energy storage system, and divides the control model into multiple sub-control models according to a preset time period; each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0252] The determination module 13 is used to determine multiple target control strategies for the electric-hydrogen hybrid energy storage system based on the distribution network information and multiple sub-control models; the sub-control models correspond one-to-one with the target control strategies.

[0253] In one embodiment, the determining module 13 includes a first determining unit, an establishing unit, and a second determining unit. The first determining unit is used to determine multiple intermediate control strategies based on distribution network information, an initial control strategy, and multiple sub-control models; the establishing unit is used to establish a global quadratic programming model based on the multiple intermediate control strategies, and solve the global quadratic programming model based on the multiple intermediate control strategies to obtain the control strategy correction amount; the second determining unit is used to determine multiple target control strategies based on the control strategy correction amount and the multiple intermediate control strategies.

[0254] In one embodiment, the first determining unit includes a first determining subunit and a second determining subunit. The first determining subunit is used to determine the augmented Lagrange model of multiple sub-control models according to the augmented Lagrange method; the second determining subunit is used to determine multiple intermediate control strategies based on distribution network information, initial control strategy, and multiple augmented Lagrange models.

[0255] In one embodiment, the second determining subunit is specifically used to construct multiple first state spaces based on the Markov decision process, according to distribution network information, initial control strategies, and multiple augmented Lagrange models; and input the multiple first state spaces into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0256] In one embodiment, the establishment unit is specifically used to determine the gradient and second-order information of each sub-control model based on multiple intermediate control strategies; and to establish a global quadratic programming model based on the multiple intermediate control strategies and the gradient and second-order information of all sub-control models.

[0257] In one embodiment, the establishment unit includes a construction subunit and an input subunit. The construction subunit is used to construct a second state space based on a Markov decision process, according to multiple intermediate control policies, the gradients and second-order information of all sub-control models, and a global quadratic programming model; the input subunit is used to input the second state space into a pre-trained second deep reinforcement learning model to obtain the control policy correction amount.

[0258] In one embodiment, the second determining unit is specifically used to modify each intermediate control strategy according to the control strategy modification amount to obtain multiple modified control strategies; if the multiple modified control strategies converge, the multiple modified control strategies are determined as the target control strategy; if the multiple modified control strategies do not converge, the multiple modified control strategies are determined as new initial control strategies, and the process returns to the steps performed by the first determining unit.

[0259] Each module in the control device of the aforementioned hybrid electric-hydrogen energy storage system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0260] In one exemplary embodiment, a computer device is provided, the internal structure of which can be as shown in the figure. Figure 11 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores distribution network information, pre-trained first and second deep reinforcement learning models, etc. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a control method for an electro-hydrogen hybrid energy storage system. The display unit of the computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0261] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0262] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0263] Obtain the distribution network information of the hybrid electric-hydrogen energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the hybrid electric-hydrogen energy storage system.

[0264] A control model for an electric-hydrogen hybrid energy storage system is established, and the control model is divided into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0265] Based on distribution network information and multiple sub-control models, several target control strategies for the electric-hydrogen hybrid energy storage system are determined; each sub-control model corresponds one-to-one with a target control strategy.

[0266] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models; establishing a global quadratic programming model based on the multiple intermediate control strategies, and solving the global quadratic programming model based on the multiple intermediate control strategies to obtain the control strategy correction amount; and determining multiple target control strategies based on the control strategy correction amount and the multiple intermediate control strategies.

[0267] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the augmented Lagrange model of multiple sub-control models according to the augmented Lagrange method; and determining multiple intermediate control strategies based on distribution network information, the initial control strategy, and the multiple augmented Lagrange models.

[0268] In one embodiment, when the processor executes the computer program, it further performs the following steps: based on the Markov decision process, constructing multiple first state spaces according to distribution network information, initial control strategies, and multiple augmented Lagrange models; inputting the multiple first state spaces into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0269] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the gradient and second-order information of each sub-control model based on multiple intermediate control strategies; and establishing a global quadratic programming model based on the multiple intermediate control strategies and the gradient and second-order information of all sub-control models.

[0270] In one embodiment, when the processor executes the computer program, it further performs the following steps: based on the Markov decision process, constructing a second state space according to multiple intermediate control policies, the gradients and second-order information of all sub-control models, and a global quadratic programming model; inputting the second state space into a pre-trained second deep reinforcement learning model to obtain the control policy correction amount.

[0271] In one embodiment, when the processor executes the computer program, it further implements the following steps: modifying each intermediate control strategy according to the control strategy modification amount to obtain multiple modified control strategies; if the multiple modified control strategies converge, determining the multiple modified control strategies as the target control strategy; if the multiple modified control strategies do not converge, determining the multiple modified control strategies as the new initial control strategy, and returning to execute the step of determining multiple intermediate control strategies according to the distribution network information, the initial control strategy, and multiple sub-control models.

[0272] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0273] Obtain the distribution network information of the hybrid electric-hydrogen energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the hybrid electric-hydrogen energy storage system.

[0274] A control model for an electric-hydrogen hybrid energy storage system is established, and the control model is divided into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0275] Based on distribution network information and multiple sub-control models, several target control strategies for the electric-hydrogen hybrid energy storage system are determined; each sub-control model corresponds one-to-one with a target control strategy.

[0276] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models; establishing a global quadratic programming model based on the multiple intermediate control strategies, and solving the global quadratic programming model based on the multiple intermediate control strategies to obtain the control strategy correction amount; and determining multiple target control strategies based on the control strategy correction amount and the multiple intermediate control strategies.

[0277] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the augmented Lagrange model of multiple sub-control models according to the augmented Lagrange method; and determining multiple intermediate control strategies based on distribution network information, the initial control strategy, and the multiple augmented Lagrange models.

[0278] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: based on the Markov decision process, constructing multiple first state spaces according to the distribution network information, the initial control strategy, and multiple augmented Lagrange models; inputting the multiple first state spaces into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0279] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the gradient and second-order information of each sub-control model based on multiple intermediate control strategies; and establishing a global quadratic programming model based on the multiple intermediate control strategies and the gradient and second-order information of all sub-control models.

[0280] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: based on the Markov decision process, constructing a second state space according to multiple intermediate control policies, the gradients and second-order information of all sub-control models, and a global quadratic programming model; inputting the second state space into a pre-trained second deep reinforcement learning model to obtain the control policy correction amount.

[0281] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: modifying each intermediate control strategy according to the control strategy modification amount to obtain multiple modified control strategies; if the multiple modified control strategies converge, determining the multiple modified control strategies as the target control strategy; if the multiple modified control strategies do not converge, determining the multiple modified control strategies as new initial control strategies, and returning to execute the step of determining multiple intermediate control strategies according to the distribution network information, the initial control strategy, and multiple sub-control models.

[0282] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0283] Obtain the distribution network information of the hybrid electric-hydrogen energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the hybrid electric-hydrogen energy storage system.

[0284] A control model for an electric-hydrogen hybrid energy storage system is established, and the control model is divided into multiple sub-control models according to a preset time period. Each sub-control model includes coupling constraints, which are the constraints between the sub-control model and the adjacent sub-control models.

[0285] Based on distribution network information and multiple sub-control models, several target control strategies for the electric-hydrogen hybrid energy storage system are determined; each sub-control model corresponds one-to-one with a target control strategy.

[0286] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining multiple intermediate control strategies based on distribution network information, initial control strategies, and multiple sub-control models; establishing a global quadratic programming model based on the multiple intermediate control strategies, and solving the global quadratic programming model based on the multiple intermediate control strategies to obtain the control strategy correction amount; and determining multiple target control strategies based on the control strategy correction amount and the multiple intermediate control strategies.

[0287] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0288] Based on the augmented Lagrange method, the augmented Lagrange models of multiple sub-control models are determined; based on distribution network information, initial control strategies, and multiple augmented Lagrange models, multiple intermediate control strategies are determined.

[0289] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: based on the Markov decision process, constructing multiple first state spaces according to the distribution network information, the initial control strategy, and multiple augmented Lagrange models; inputting the multiple first state spaces into a pre-trained first deep reinforcement learning model to obtain multiple intermediate control strategies.

[0290] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the gradient and second-order information of each sub-control model based on multiple intermediate control strategies; and establishing a global quadratic programming model based on the multiple intermediate control strategies and the gradient and second-order information of all sub-control models.

[0291] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: based on the Markov decision process, constructing a second state space according to multiple intermediate control policies, the gradients and second-order information of all sub-control models, and a global quadratic programming model; inputting the second state space into a pre-trained second deep reinforcement learning model to obtain the control policy correction amount.

[0292] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: modifying each intermediate control strategy according to the control strategy modification amount to obtain multiple modified control strategies; if the multiple modified control strategies converge, determining the multiple modified control strategies as the target control strategy; if the multiple modified control strategies do not converge, determining the multiple modified control strategies as new initial control strategies, and returning to execute the step of determining multiple intermediate control strategies according to the distribution network information, the initial control strategy, and multiple sub-control models.

[0293] It should be noted that the data involved in this application (including but not limited to data used for analysis, data stored, data displayed, etc.) are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0294] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0295] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0296] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for regulating an electro-hydrogen hybrid energy storage system, characterized in that, The method includes: Obtain the distribution network information of the electric-hydrogen hybrid energy storage system; the distribution network information includes the source-load data, operating parameters, and distribution network topology information of the electric-hydrogen hybrid energy storage system. A control model for the electric-hydrogen hybrid energy storage system is established, and the control model is divided into multiple sub-control models according to a preset time period; each sub-control model includes coupling constraints, which are constraints between the sub-control model and adjacent sub-control models. Based on the power distribution network information and the multiple sub-control models, multiple target control strategies for the electric-hydrogen hybrid energy storage system are determined; each sub-control model corresponds one-to-one with a target control strategy.

2. The control method according to claim 1, characterized in that, The determination of multiple target control strategies for the electric-hydrogen hybrid energy storage system based on the distribution network information and the multiple sub-control models includes: Based on the distribution network information, the initial control strategy, and the multiple sub-control models, multiple intermediate control strategies are determined; A global quadratic programming model is established based on the multiple intermediate control strategies, and the global quadratic programming model is solved based on the multiple intermediate control strategies to obtain the control strategy correction amount; Based on the adjustment amount of the control strategy and the plurality of intermediate control strategies, the plurality of target control strategies are determined.

3. The control method according to claim 2, characterized in that, The step of determining multiple intermediate control strategies based on the distribution network information, the initial control strategy, and the multiple sub-control models includes: The augmented Lagrange model of the multiple sub-regulation models is determined according to the augmented Lagrange method; Based on the distribution network information, the initial control strategy, and multiple augmented Lagrange models, multiple intermediate control strategies are determined.

4. The control method according to claim 3, characterized in that, The step of determining multiple intermediate control strategies based on the distribution network information, the initial control strategy, and multiple augmented Lagrange models includes: Based on the Markov decision process, multiple first state spaces are constructed according to the distribution network information, the initial control strategy, and multiple augmented Lagrange models; The multiple first state spaces are input into a pre-trained first deep reinforcement learning model to obtain the multiple intermediate control strategies.

5. The control method according to claim 2, characterized in that, The step of establishing a global quadratic programming model based on the multiple intermediate control strategies includes: Based on the multiple intermediate control strategies, determine the gradient and second-order information of each sub-control model; A global quadratic programming model is established based on the multiple intermediate control strategies and the gradient and second-order information of all the sub-control models.

6. The control method according to claim 5, characterized in that, The step of solving the global quadratic programming model based on the multiple intermediate control strategies to obtain the control strategy correction includes: Based on the Markov decision process, a second state space is constructed according to the multiple intermediate control strategies, the gradients and second-order information of all the sub-control models, and the global quadratic programming model. The second state space is input into a pre-trained second deep reinforcement learning model to obtain the adjustment amount of the regulation policy.

7. The control method according to claim 2, characterized in that, The step of determining the plurality of target control strategies based on the control strategy correction amount and the plurality of intermediate control strategies includes: Each intermediate control strategy is modified according to the control strategy correction amount to obtain multiple modified control strategies; If the plurality of corrective control strategies converge, the plurality of corrective control strategies shall be determined as the target control strategy. If the multiple modified control strategies fail to converge, the multiple modified control strategies are determined as new initial control strategies, and the process returns to the step of determining multiple intermediate control strategies based on the distribution network information, the initial control strategies, and the multiple sub-control models.

8. A control device for an electro-hydrogen hybrid energy storage system, characterized in that, The device includes: The acquisition module is used to acquire the distribution network information of the electric-hydrogen hybrid energy storage system. The distribution network information includes the source-load data, operating parameters and distribution network topology information of the electric-hydrogen hybrid energy storage system. A module is established to create a control model for the electric-hydrogen hybrid energy storage system, and to divide the control model into multiple sub-control models according to a preset time period; each sub-control model includes coupling constraints, which are constraints between the sub-control model and adjacent sub-control models. The determination module is used to determine multiple target control strategies for the electric-hydrogen hybrid energy storage system based on the distribution network information and the multiple sub-control models; the sub-control models correspond one-to-one with the target control strategies.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method according to any one of claims 1 to 7.

Images