RLDEA-based optimization method for wind power-hydrogen energy coordinated power collection system

Abstract

The present application belongs to the technical field of offshore wind power power collection system planning, and discloses a wind power-hydrogen energy collaborative power collection system optimization method, device, equipment and medium based on RLDEA. With the minimum life cycle cost of the power collection system and the maximum hydrogen energy utilization rate of local energy utilization as the target, a multi-dimensional constraint system containing topology constraints, submarine cable constraints, operation constraints, hydrogen energy collaboration constraints and wake constraints is constructed. The wind power output and hydrogen energy demand are modeled with double uncertainty, and typical wind-hydrogen coupling scenarios are obtained through historical scenario clustering. A distributed robust planning model based on the RLDEA algorithm is established, the differential evolution parameters are dynamically adjusted combined with reinforcement learning, the power collection system clustering partition, topology connection, submarine cable selection and hydrogen energy access point synchronous optimization are realized, and the optimal planning scheme considering economy, reliability and hydrogen energy adaptability is obtained by using the column and constraint generation algorithm.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power collection system planning technology, specifically to an optimization method, apparatus, equipment, and medium for a wind power-hydrogen energy collaborative collection system based on RLDEA. Background Technology

[0002] Offshore wind power collection systems are the core hub connecting wind turbines and substations, and their planning quality directly determines the investment cost, operational efficiency, and maintenance difficulty of wind farms. With the expansion of offshore wind farms to the scale of hundreds of turbines and the rapid development of on-site hydrogen energy utilization technology, collection system planning faces new challenges: Traditional collection system planning often uses heuristic algorithms (such as genetic algorithms and particle swarm optimization), which are prone to getting trapped in local optima and struggle to simultaneously optimize the three core decisions of clustering partitioning, topology connectivity, and submarine cable selection, leading to high collection costs; precise algorithms such as mixed-integer linear programming have extremely low solution efficiency in large-scale wind farms, making it difficult to meet the real-time planning requirements of engineering projects, and lacking flexibility in handling constraints such as topology radiation and cable crossing avoidance; existing planning methods often ignore the uncertainty of wind power output and the wake effect. The existing plans, while addressing the distributed access requirements for on-site hydrogen energy utilization, have resulted in issues such as excessive voltage drop, submarine cable overload, and poor adaptability to power fluctuations in hydrogen energy utilization during actual operation. On-site hydrogen energy utilization requires wind turbines to directly supply power to distributed electrolyzers, necessitating a power collection system that adapts to this dual characteristic of "distributed power supply and centralized collection." However, current plans lack corresponding power allocation constraints and access point optimization mechanisms, leading to low on-site hydrogen energy utilization rates and high wind curtailment rates. Furthermore, insufficient coordination and optimization of submarine cable type selection, topology, and hydrogen energy access points have resulted in a failure to fully balance construction costs, power loss costs, and the benefits of on-site hydrogen energy utilization, leading to poor economic efficiency throughout the entire lifecycle. The core objective of this invention is to minimize the discounted cost of the power collection system throughout its entire lifecycle, simultaneously meeting the requirements for power adaptation, access point optimization, and maximizing energy utilization for on-site hydrogen energy utilization. This addresses the problems of premature convergence in traditional heuristic algorithms, dimensionality explosion in precise algorithms, and poor constraint coupling handling.

[0003] This invention proposes an optimization method, device, equipment, and medium for wind power-hydrogen energy collaborative collection systems based on RLDEA. It is applicable to clustering and zoning, radial topology design, and submarine cable selection for large-scale offshore wind farm collection systems. It also adapts to the distributed access requirements for on-site hydrogen energy utilization, taking into account the economy, operational reliability, and comprehensive energy utilization rate of the collection system. It can be widely applied to large-scale offshore wind power projects in deep and far sea areas. Summary of the Invention

[0004] This invention provides an optimization method, device, equipment, and medium for wind power-hydrogen energy collaborative collection systems based on RLDEA, thereby effectively solving the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an optimization method for a wind power-hydrogen energy collaborative power collection system based on RLDEA, comprising the following steps:

[0006] S1: Establish a core objective function and a multi-dimensional constraint system, construct a life-cycle discounted cost objective function that includes the time value of money and integrate the benefits of local utilization of hydrogen energy, and simultaneously build a four-dimensional constraint system that integrates electricity, hydrogen energy, dual uncertainties, and equipment linkage to anchor the optimization direction and operating boundary.

[0007] S2: Modeling the dual uncertainty of wind and hydrogen, using the K-means clustering algorithm to reduce the dimensionality of massive original wind and hydrogen scenarios, generating typical coupled scenarios and joint probabilities, solving the problem of dimensionality explosion in model solving, and significantly reducing the computational load.

[0008] S3: Construct a planning model for a distributed bar collector system based on the RLDEA algorithm, integrate the advantages of reinforcement learning and differential evolution algorithms, achieve simultaneous optimization of four-dimensional decision-making, and improve the scheme's resistance to uncertainty interference by combining distributed bar theory.

[0009] S4: Model linearization processing. For nonlinear parts such as power loss quadratic terms and uncertainty norm constraints, auxiliary variables are introduced to transform them into a directly calculable mixed integer linear programming model.

[0010] S5: The column and constraint generation algorithm solves the linearized model by splitting it into the main investment decision problem and the sub-problem of operating cost. It generates pruning constraints through iterative interaction, and outputs the optimal planning scheme after satisfying the convergence criterion, forming a complete technical closed loop.

[0011] First, we construct a life-cycle discounted cost objective function incorporating the time value of money, while also considering the benefits of local hydrogen energy utilization. The objective function expression is as follows:

[0012]

[0013] in The present value is the discounted net cost over the entire lifecycle of the power collection system, expressed in ten thousand yuan. =25 is the design life of the collector system. =7.5%~8% is the discount rate for funds. For initial construction costs, , The first Annual power loss cost and operation and maintenance cost The hydrogen energy value coefficient is set at 1.2~1.3 yuan / kWh. The energy conversion efficiency for hydrogen production in an electrolyzer is taken as 0.85~0.86. For the first Total annual electricity generated from local hydrogen utilization This is the discount rate for funds, ranging from 7.5% to 8%.

[0014] To clarify the calculation logic of each component of the objective function, it is broken down into its constituent parts. The initial construction cost includes the investment in the current collection submarine cable, the booster station, and the supporting equipment for hydrogen energy access. The calculation formula is as follows:

[0015]

[0016]

[0017] in This represents the node pair at the connection point of the wind turbine electrolytic cell in the current collection system. This represents the set of all potential connection edges in the collector system. The cost per unit length of type b submarine cable is expressed in ten thousand yuan / km. For nodes With nodes The distance between them, in km. For nodes With nodes The cable connection status is indicated by 1 for a connected cable and 0 for a disconnected cable. The cost of construction and expansion of the substation is expressed in ten thousand yuan. It is the set of electrolytic cell connection points. For the first The cost of supporting equipment for each electrolytic cell access point is in ten thousand yuan. Representing the The construction cost of the connection cable to each electrolytic cell, Cost per unit length of access cable, This refers to the length of the access cable.

[0018] The annual power loss cost needs to consider the impact of hydrogen diversion on the transmission power of the power collection branches, while also taking into account the annual fluctuations in electricity prices. The calculation formula is as follows:

[0019]

[0020]

[0021] in This represents the annual operating hours, with a value of 8760 hours. For the first Annual on-grid electricity price, For the first The resistance per unit length of this type of submarine cable, expressed in Ω / km. For the first Annual collection branch Total transmission power, in MW. For the first Annual collection branch The power diverted to the electrolytic cell, measured in MW. The rated voltage of the current collection system, For power factor, Based on the benchmark annual electricity price, =2% is the annual growth rate of electricity prices.

[0022] The annual benefits of on-site hydrogen energy utilization are directly related to the scale of electricity utilization and hydrogen production efficiency. The calculation formula is as follows:

[0023]

[0024] in For the first The total economic benefit of the hydrogen energy system within an optimization cycle is a quantified measure of the comprehensive economic value brought about by on-site hydrogen production from wind power. It is the core benefit item in the optimization objective used to reduce system costs and guide the optimization of hydrogen energy access points, and its unit is usually yuan / cycle. For the first Year Hourly collector branch The power diverted to the electrolytic cell, measured in MW. The time step is 1 hour. The hydrogen energy value coefficient; The energy conversion efficiency of hydrogen production in an electrolyzer.

[0025] To further align with actual hydrogen production capacity targets in engineering projects, a supplementary formula relating on-site hydrogen utilization to hydrogen production capacity is provided:

[0026]

[0027] In the formula The total amount of hydrogen produced locally in year t, in kg. =0.0397 is the efficiency coefficient for hydrogen production through water electrolysis. =39.4 is the lower calorific value of hydrogen. This formula can be used to directly calculate the total amount of hydrogen produced locally each year.

[0028] Furthermore, considering the construction of a four-dimensional constraint system, and taking into account the core constraints of the power collection system to ensure its safe and stable operation, the first constraint is the topological radial constraint. This constraint ensures that the power collection system is a loop-free radial structure with the step-up substation as the root node, expressed as:

[0029]

[0030]

[0031] in For the set of all nodes in the collector system, This is a set of wind turbine nodes. This represents the direct connection status between the booster station and fan i, where 1 indicates connection and 0 indicates no connection. This represents the maximum number of feeders in the collector system, with a value ranging from 10 to 15.

[0032] Secondly, there is the constraint on the current carrying capacity of the submarine cable. Both the current transmission current of the current collector branch and the hydrogen shunt current must be limited, and the hydrogen shunt current must not exceed 40% of the branch's rated current. The constraint formula is as follows:

[0033]

[0034]

[0035] in For the collector system branch The total transmitted active power refers to the power transmitted from the node. Flow to Node The total active power, branch road Active power supplied to the hydrogen energy system (electrolyzer / hydrogen production station) For the first The rated current carrying capacity of this type of submarine cable, in amperes (A). Power factor; This is the rated voltage of the power grid.

[0036] Voltage drop constraints need to consider both the main power grid and the branch lines connecting to the electrolytic cells separately, ensuring that the voltage drop of both types of lines is within the allowable range. The constraint formula is:

[0037]

[0038]

[0039] Among them, reactive power It can be calculated using active power and power factor, using the following formula:

[0040]

[0041]

[0042] branch road Voltage drop; For the first Voltage drop in the hydrogen energy access branch; For the first The reactance per unit length of this type of submarine cable, expressed in Ω / km. The maximum allowable voltage drop of the collector system. For the first The input power of each electrolytic cell, measured in MW. For the first The resistance per unit length of the cable connecting to each electrolytic cell, expressed in Ω / km. For the first The length of the cable connecting each electrolytic cell is expressed in km. This is the rated voltage of the electrolytic cell. This represents the maximum permissible voltage drop of the branch line connected to the electrolytic cell.

[0043] Clustering partitioning constraints are used to achieve regionalized management of the connection points of wind turbines and electrolyzers, ensuring balanced power load within a single area. The constraint formula is as follows:

[0044]

[0045]

[0046] in For nodes Belongs to the The identifiers for each cluster region are 1 (representing a cluster) and 0 (representing a cluster not representing a cluster). 'n' represents the number of cluster regions, ranging from 5 to 6. For the first A collection of wind turbines in a clustered region For wind turbine Rated power, in MW. This represents the maximum total power allowed in a single cluster region, with a value ranging from 120 to 150 MW.

[0047] To ensure the stability and efficiency of on-site hydrogen energy utilization, the first constraint is the operating power constraint of the electrolyzer. This constraint prevents the electrolyzer from operating within an unsafe power range. The formula is as follows:

[0048]

[0049] in This represents the minimum operating power of electrolytic cell k. This represents the maximum operating power of electrolytic cell k. The value represents the operating status of electrolytic cell k in the τth hour, where 1 indicates operation and 0 indicates shutdown.

[0050] Secondly, there is a constraint on the proportion of on-site hydrogen energy utilization, ensuring that the hydrogen energy utilization rate of each cluster area is not lower than a set lower limit. The constraint formula is as follows:

[0051]

[0052] in The minimum on-site utilization rate of hydrogen energy is set at 0.15 to 0.18.

[0053] To address the dual uncertainties of wind and hydrogen demand, a piecewise function model is used to calculate the wind turbine output based on the wind speed range. The formula is as follows:

[0054]

[0055] in Wind speed, in m / s. The cut-in wind speed of the fan. This refers to the cut-off velocity of the fan. The rated power of the wind turbine is expressed in MW. The rated wind speed of the fan.

[0056] Hydrogen demand is allowed to fluctuate within a reasonable range of forecasts, with the constraint formula as follows:

[0057]

[0058] in The predicted power output for hydrogen energy demand in electrolyzer k at hour τ is given, in MW. The actual hydrogen energy demand of electrolyzer k in hour τ is expressed in MW.

[0059] Considering the linkage constraints of equipment operation, to ensure that the operating status of the fan, current collection branch and electrolytic cell are consistent, and to avoid situations where the equipment is unloaded or the power supply is interrupted, the constraint formula is:

[0060]

[0061]

[0062] in This represents the operating status of the current collector feeder corresponding to electrolytic cell k at hour τ, where 1 indicates operation and 0 indicates shutdown. The set of collector branches that supply power to electrolytic cell k.

[0063] Furthermore, considering the modeling of wind-hydrogen dual uncertainties, this step employs the K-means clustering algorithm to cluster the massive amount of original wind-hydrogen scenarios, generating typical wind-hydrogen coupled scenarios. The core objective of clustering is to minimize the sum of squared distances between scenarios within each cluster and the cluster center. The objective function is:

[0064]

[0065] in The objective function value for clustering. This represents the number of clusters, with a value between 5 and 6. For the first A collection of scenarios for clustering. For the scene The total wind power output, in MW. The total hydrogen energy demand in scenario h is expressed in MW. For the first The center of each cluster.

[0066] During the clustering iteration process, the center coordinates of each cluster need to be continuously updated to ensure that the clustering effect converges. The formula for updating the cluster centers is:

[0067]

[0068] in For the first The number of scenes within each cluster.

[0069] Furthermore, we consider constructing a planning model for a distributed bar collector system based on the RLDEA algorithm, integrating the advantages of reinforcement learning and differential evolution algorithms to achieve adaptive adjustment of algorithm parameters and synchronous optimization across multiple decision dimensions.

[0070] The reinforcement learning module is responsible for dynamically adjusting the core parameters of the differential evolution algorithm. Its state space encompasses key indicators such as power collection system cost, constraint satisfaction rate, hydrogen energy utilization rate, and algorithm parameters, and its expression is:

[0071]

[0072] in For the first The state vector of the iteration, No. The total cost of the iterative current collector system, in ten thousand yuan. For the first The constraint satisfaction rate of the iteration. For the first The local utilization rate of hydrogen energy through iterative iterations; No. The probability of differential evolution mutation in each iteration. No. Differential evolution crossover probability in iterations. For the first The number of clustered regions in each iteration.

[0073] The action space is the adjustment step size for each parameter. Simultaneously, the adjustment range of the parameters must be limited to ensure that they fluctuate within a reasonable range. The formula is:

[0074]

[0075]

[0076]

[0077] in For the first The action vector of the iteration, This is the adjustment step size for the mutation probability, with a value ranging from -0.2 to 0.2. This is the adjustment step size for the crossover probability, with a value ranging from -0.2 to 0.2. This is the adjustment step size for the number of cluster regions, with a value ranging from -1 to 1. The minimum mutation probability is 0.3. This represents the maximum mutation probability, with a value of 0.9. The minimum crossover probability is 0.4. This represents the maximum crossover probability, with a value of 0.8.

[0078] The reward function of reinforcement learning comprehensively considers the cost reduction rate, constraint satisfaction rate, and hydrogen energy utilization rate, thereby guiding the algorithm to iterate towards the optimal solution. The formula is as follows:

[0079]

[0080] in For the first The reward value of each iteration, , , These are the weighting coefficients.

[0081] The differential evolution module is responsible for achieving four-dimensional synchronous optimization of power collection system clustering and partitioning, topology connection, submarine cable selection, and hydrogen energy access point. First, each decision-making entity is encoded in four dimensions, using the following formula:

[0082]

[0083] in For the first The encoding vector of each decision-making individual. For cluster partitioning encoding, For topology connection encoding, For submarine cable model codes, Encoding for hydrogen energy access points.

[0084] Then, mutation, crossover, and selection operations are performed sequentially to generate a new generation of decision-making individuals. The mutation operation formula is as follows:

[0085]

[0086] Let be the mutation vector of the t-th iteration. For randomly selected individual indexes, i.e. These are three distinct individual vectors randomly selected from the population. This is the clustering number adjustment factor, with a value of 0.1.

[0087] The formula for crossover operation is:

[0088]

[0089] in Let be the cross vector of the t-th iteration; A random number between 0 and 1; The dimension index of the encoded vector; For the first Individual, the first A vector of mutated individuals of a dimensional variable; For the first Individual, the first The original individual vector of the dimension variable.

[0090] The selection operation employs a greedy strategy, retaining individuals with better objective function values. The formula is as follows:

[0091]

[0092] in For the first The individual in the first The vector of the experimental individuals of the generation; For the first The individual in the first The original individual vector of the generation.

[0093] Furthermore, considering the impact of wind and hydrogen uncertainties on the planning scheme, a sub-Bruker optimization model is constructed based on the above algorithm framework. The model uses investment decisions as optimization variables and also considers the operating cost under the worst-case scenario. The formula is:

[0094]

[0095] in For investment decision variables; The feasible region for investment decision variables; The joint probability distribution for wind-hydrogen coupling scenarios; It is the set of uncertainties in the joint probability distribution; For running decision variables; For a given investment decision The feasible region of the decision variables under running conditions; Based on joint probability distribution The expected operation.

[0096] Furthermore, considering model linearization, since the above model contains nonlinear components such as the quadratic term of power loss and uncertainty norm constraints, it cannot be directly calculated using a linear programming solver. Therefore, the model needs to be linearized. For the quadratic term of submarine cable power loss, auxiliary variables are introduced. Transform it into a linear constraint, the formula is:

[0097]

[0098]

[0099]

[0100] in For collector branch The transmission current, measured in amperes (A). This represents the minimum value of the current transmitted through the collector branch. This represents the maximum value of the current transmitted through the collector branch, which is taken as the rated current carrying capacity of the submarine cable. The power loss of collector branch (i,j) is expressed in MW.

[0101] For the 1-norm constraint in the uncertainty set, its linear expansion is transformed into an inequality constraint, as shown in the formula:

[0102]

[0103] in Let be the initial joint probability of the wind-hydrogen coupling scenario. The probability deviation limit for the 1-norm is 0.09 to 0.1.

[0104] Furthermore, considering the algorithmic solution, a generative algorithm is used to solve the linearized model. This algorithm breaks down the model into a main problem and subproblems, gradually converging to the optimal solution through iterative interaction. The main problem is a mixed-integer linear programming problem used to optimize the investment decision variables of the power collection system. The model expression is:

[0105]

[0106]

[0107] in Let R be the upper bound of the operating cost in year t, expressed in ten thousand yuan, and R be the set of real numbers.

[0108] The subproblem is a linear programming problem. Given the investment decision in the main problem, we need to solve for the worst-case operating cost. The model expression is:

[0109]

[0110] After each iteration, the optimal solution of the subproblem is fed back to the main problem as a pruning constraint. Iteration stops when the relative difference between the upper and lower bounds of the main problem is less than the convergence threshold. The convergence criterion is:

[0111]

[0112] in Let be the upper bound of the objective function. This is the lower bound of the objective function. The convergence threshold is set to a value of 0.008 to 0.01.

[0113] The present invention also includes an optimization device for a wind power-hydrogen energy co-collection system based on RLDEA, using the method described above, including:

[0114] Establish a core objective function and a multi-dimensional constraint system, construct a life-cycle discounted cost objective function that includes the time value of money, and integrate the benefits of local utilization of hydrogen energy to anchor the optimization direction and operational boundary.

[0115] To model the dual uncertainties of wind and hydrogen, the K-means clustering algorithm is used to reduce the dimensionality of massive original wind and hydrogen scenarios, generating typical coupled scenarios and joint probabilities, which significantly reduces the computational load.

[0116] A planning model for a distributed bar collector system based on the RLDEA algorithm is constructed, which integrates the advantages of reinforcement learning and differential evolution algorithm to improve the scheme's ability to resist uncertainty interference.

[0117] The model is linearized by introducing auxiliary variables and other methods to form a directly computable mixed-integer linear programming model.

[0118] The generative algorithm solves the linearized model by breaking it down into the main investment decision problem and the sub-problem of operating cost. It generates pruning constraints through iterative interaction, and outputs the optimal planning scheme after satisfying the convergence criterion, thus forming a complete technical closed loop.

[0119] The present invention also includes a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as described above.

[0120] The present invention also includes a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described above.

[0121] Compared with the prior art, the present invention has the following significant advantages:

[0122] The economic efficiency throughout the entire life cycle is significantly improved. Through global optimization using the RLDEA algorithm, the net cost discounted throughout the life cycle is reduced by 20% compared to the traditional genetic algorithm. At the same time, the adaptability of hydrogen energy for local utilization is excellent. Through access point optimization and power balance constraints, the local utilization rate of hydrogen energy reaches 22.8%. The solution efficiency is greatly improved, and the solution time is shortened by 72% compared to the traditional algorithm. It has strong robustness. In the wind-hydrogen coupling scenario, the wind curtailment rate is only 1.0% and the risk of submarine cable overload is 0 under extreme scenarios. Attached Figure Description

[0123] Figure 1 Here is the overall flowchart of the optimization method for wind power-hydrogen energy synergistic collection system based on RLDEA;

[0124] Figure 2 This is a radial topology diagram of the wind-hydrogen synergistic power collection system in Example 1;

[0125] Figure 3 This is a radial topology diagram of the wind-hydrogen synergistic power collection system in Example 2;

[0126] Figure 4 Verification diagram showing the compliance of submarine cable current carrying capacity and voltage drop constraints;

[0127] Figure 5 This is a comparison chart of the iterative convergence of RLDEA and the traditional algorithm in Example 1;

[0128] Figure 6 This is a comparison chart of the iterative convergence of RLDEA and the traditional algorithm in Example 2;

[0129] Figure 7 Diagram of a distributed access architecture for local utilization of hydrogen energy;

[0130] Figure 8 Verification diagram of dynamic power matching between current collector branch and electrolyzer (24h);

[0131] Figure 9 This is a schematic diagram of the structure of a computer device. Detailed Implementation

[0132] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0133] like Figure 1 As shown: The optimization method for wind power-hydrogen energy synergistic power collection system based on RLDEA includes the following steps:

[0134] S1: Establish a core objective function and a multi-dimensional constraint system, construct a full life-cycle discounted cost objective function that includes the time value of money and integrate the benefits of local utilization of hydrogen energy, and simultaneously build a four-dimensional constraint system that integrates electricity, hydrogen energy, dual uncertainties, and equipment linkage to anchor the optimization direction and operating boundary;

[0135] S2: Wind-hydrogen dual uncertainty modeling, using K-means clustering algorithm to reduce the dimensionality of massive original wind-hydrogen scenarios, generating typical coupled scenarios and joint probabilities, solving the problem of dimensionality explosion in model solution, and significantly reducing the computational load;

[0136] S3: Construct a planning model for a distributed bluish bar collector system based on the RLDEA algorithm, integrate the advantages of reinforcement learning and differential evolution algorithm, achieve synchronous optimization of four-dimensional decision-making, and improve the scheme's resistance to uncertainty interference by combining distributed bluish bar theory;

[0137] S4: Model linearization processing: For nonlinear parts such as power loss quadratic terms and uncertainty norm constraints, the model is transformed by introducing auxiliary variables to form a directly computable mixed integer linear programming model.

[0138] S5: The column and constraint generation algorithm solves the linearized model by splitting it into the main investment decision problem and the sub-problem of operating cost. It generates pruning constraints through iterative interaction, and outputs the optimal planning scheme after satisfying the convergence criterion, forming a complete technical closed loop.

[0139] The core objective of minimizing the life-cycle cost of the power collection system, while maximizing the energy utilization rate of hydrogen energy on-site, includes:

[0140] The objective function is to achieve the optimal balance between the total life-cycle cost of the power collection system and the benefits of on-site hydrogen utilization, expressed as:

[0141]

[0142] Among them: construction cost of the power collection system :

[0143]

[0144] In the formula: For the first Cost per unit length of submarine cable of various types (ten thousand yuan / km); For nodes and Distance (km); For nodes and Submarine cable connection status ( =1 indicates a connection, 0 indicates the opposite); Cost of construction and expansion of the substation (ten thousand yuan); This refers to the set of electrolytic cell connection points. For the first Cost of supporting equipment (rectifier, switch) for each electrolytic cell access point (ten thousand yuan); Let represent the set of all potential connection edges in the collector system, i.e.:

[0145]

[0146] in It is a collection of all nodes in the power collection system, including three core types of nodes: wind turbine nodes, offshore substation nodes, and electrolyzer (hydrogen production station) access nodes;

[0147] Represents a node With nodes An undirected potential connection edge between two nodes represents the engineering possibility of laying a submarine cable between the two nodes;

[0148] Power loss cost of collector system :

[0149]

[0150] In the formula: =8760 represents the number of operating hours per year; The on-grid electricity price (RMB / kWh); The resistance per unit length of the b-type submarine cable (Ω / km); The total power (MW) of the collector branch (i,j); The power (MW) supplied locally to the electrolytic cell for the current collector branch (i,j); Rated voltage (kV); =0.95 is the power factor;

[0151] Operation and maintenance costs of power collection systems :

[0152]

[0153] In the formula: For the first Annual maintenance cost per unit length of submarine cable (ten thousand yuan / km·year). The annual operation and maintenance cost of the substation (ten thousand yuan). For the first Annual operation and maintenance cost per unit capacity of electrolytic cell (ten thousand yuan / MW·year); For the first Rated power (MW) of each electrolytic cell;

[0154] Benefits of on-site utilization of hydrogen energy: =1.2 is the hydrogen energy value coefficient (yuan / kWh); =0.85 represents the energy conversion efficiency of hydrogen production in an electrolyzer; The total annual on-site hydrogen energy utilization (MWh) is expressed as:

[0155]

[0156] In the formula: For a moment collector branch Power supplied to the electrolyzer on-site (MW); =1h is the time step.

[0157] To ensure the engineering feasibility and robustness of the planning scheme, and to adapt to the fluctuation characteristics of the initial dual-uncertainty modeling, the established multi-dimensional optimization constraints specifically include the following: all constraints match the boundary of the uncertainty set obtained from the modeling, ensuring that the constraint requirements can cover the fluctuation range of the wind-hydrogen dual sources.

[0158] Core constraints of the power collection system: topological radiality constraint, submarine cable current carrying capacity constraint, voltage drop constraint, cable crossing avoidance constraint, and clustering partitioning constraint;

[0159] Coordination constraints for local hydrogen energy utilization: power constraints for electrolyzer access, proportion constraints for local hydrogen energy utilization, power balance constraints between power collection branch and electrolyzer, and linkage constraints between electrolyzer start-up / shutdown and power collection system operation.

[0160] Dual uncertainty adaptation constraints: power balance constraints and robustness constraints in wind-hydrogen coupling scenarios;

[0161] Equipment operating status constraints: start-stop linkage constraints between the fan-collector branch-electrolyte cell and the matching constraints between the submarine cable and the electrolyte cell access point.

[0162] As the foundation for the safe operation of the system, and to adapt to the fluctuation requirements of dual uncertainty modeling, the core constraints of the power collection system specifically include the following: the limit values ​​of each constraint are set in conjunction with the wind-hydrogen power fluctuation range obtained from the modeling, ensuring that the system can still operate safely even under extreme fluctuation scenarios:

[0163] Topological radial constraints:

[0164]

[0165]

[0166]

[0167] In the formula: For the set of all nodes in the collector system, For wind turbine nodes; For the set of nodes of the booster station; For booster stations and wind turbines The direct connection state; This represents the maximum number of feeders; This refers to the number of electrolytic cell connection points;

[0168] Submarine cable current carrying capacity constraints:

[0169]

[0170]

[0171] In the formula: The rated current carrying capacity (A) of the type b submarine cable; For collector branch The maximum allowable current (A) allocated to the electrolytic cell satisfies ≤0.4 ;

[0172] Voltage drop constraint:

[0173]

[0174]

[0175] In the formula: It is a node arrive The total voltage drop of the submarine cable includes the voltage loss caused by the combined effect of wind turbine power generation and hydrogen production load power; It is the voltage drop component under the sole action of hydrogen production load, characterizing the impact of electrolyzer connection on submarine cable voltage drop, and is used to quantify the disturbance of hydrogen production load on the voltage of current collection system. The reactance per unit length of the b-type submarine cable (Ω / km); For collector branch reactive power (Mvar); =5% is the maximum allowable voltage drop of the collector system; The distance (km) from the electrolytic cell access point to the collector branch node; Resistance per unit length of the cable connecting to the electrolytic cell (Ω / km); =0.69kV is the rated voltage of the electrolytic cell; =3% is the maximum allowable voltage drop when the electrolytic cell is connected;

[0176] Cable crossing avoidance constraints:

[0177]

[0178] In the formula: Represents the 0-1 connection decision variables between two potential submarine cable edges; This represents two intersecting submarine cable edges, signifying a potential connection between two node pairs. These are pre-determined as intersecting edge pairs through planar geometric cross product operations. It is a subset of the set E of all potential edges in the current collector system, including all submarine cable edge pairs that will cross according to geometric determination, and is the scope of this constraint.

[0179] Clustering partitioning constraints:

[0180]

[0181]

[0182] In the formula: For nodes (Fan / Electrolyzer Connection Point) belongs to the first The identifier of each cluster region (1 indicates belonging, 0 indicates not); n is the number of cluster regions; each cluster region has at least 3 fans and at most 2 electrolytic cell access points.

[0183] To achieve efficient on-site hydrogen energy utilization and simultaneously address the wind-hydrogen power fluctuation characteristics under dual uncertainty modeling, the collaborative constraints for on-site hydrogen energy utilization specifically include constraints set in conjunction with typical wind-hydrogen coupling scenarios obtained from modeling, ensuring that hydrogen energy utilization can adapt to the random fluctuations in wind power output.

[0184] Electrolytic cell access power constraints:

[0185]

[0186]

[0187] In the formula: No. Total power of the supporting equipment at each electrolytic cell access point The minimum / maximum operating power of the electrolytic cell; For a moment No. Operating status of each electrolytic cell (1 in operation, 0 out of operation); For a moment No. Total input power (MW) of each electrolytic cell; To give the first A collection of current collector branches supplying power to each electrolytic cell;

[0188] Constraints on the proportion of hydrogen energy used locally:

[0189]

[0190] In the formula: For a moment No. The total wind power (MW) of the clustered regions corresponding to each electrolytic cell; =0.15 is the minimum on-site utilization rate of hydrogen energy;

[0191] Current collector branch - electrolytic cell power balance constraint:

[0192]

[0193] In the formula: For a moment collector branch The wind turbine input power (MW); For a moment collector branch Power loss (MW);

[0194] Electrolytic cell start-up / shutdown and current collection system operation linkage constraints:

[0195]

[0196]

[0197] In the formula: For a moment No. The operating status of the current collector feeder corresponding to each electrolytic cell (1 is running, 0 is out of service).

[0198] To address the dual uncertainties of wind and hydrogen and achieve efficient optimization, the core of the sub-blob rod current collector system planning model based on the RLDEA algorithm is to establish a linkage mechanism between the modeling results and the algorithm solution. Specifically, all modules of the algorithm are adapted to the output of the previous dual-uncertainty modeling, ensuring that the algorithm can accurately utilize the modeling information to achieve robust optimization.

[0199] Reinforcement learning module:

[0200] State space:

[0201]

[0202] in The construction cost corresponds to the submarine cable construction cost and the equipment cost of the booster station / hydrogen production station in the objective function, representing the current economic investment level of the topology; The network loss cost corresponds to the submarine cable resistance loss and equipment operating loss in the objective function, and characterizes the system operating efficiency. For submarine cable current, the first The actual current (A) on the submarine cable characterizes the cable load factor; For voltage drop, for the first Voltage deviation of a submarine cable. Characterizes electrical stability; It is a cluster label that identifies the cluster region to which each node belongs and represents the topological partitioning structure. This is the discount factor / penalty coefficient, an adaptive parameter within the reinforcement learning algorithm. It represents the balance between exploration and exploitation. For a moment The total on-site hydrogen energy utilization power (MW) characterizes the level of on-site hydrogen energy utilization. This is the core indicator of this invention, directly determining whether energy utilization efficiency is maximized.

[0203] Action space:

[0204]

[0205] in Let t be the action vector of the t-th iteration. This is the adjustment step size for the mutation probability, with a value ranging from -0.2 to 0.2. This is the adjustment step size for the crossover probability, with a value ranging from -0.2 to 0.2. This is the adjustment step size for the number of cluster regions, with a value ranging from -1 to 1. Adjust the step size for the proportion of hydrogen energy used locally.

[0206] Reward function:

[0207]

[0208] In the formula: =0.6、 =0.2、 =0.1、 =0.1 is the weighting coefficient; The total output of the wind turbine at time t (MW); The number of feasible solutions; The total number of solutions; The total lifecycle cost of the power collection system at time t includes the cost of submarine cable construction, network loss, and operation and maintenance, excluding the benefits of on-site utilization of hydrogen energy.

[0209] Differential Evolution Module:

[0210] Individual code:

[0211]

[0212] in Optimize the coding for the electrolytic cell connection point. For topology connection encoding, representing the first In the topology corresponding to each individual, the nodes With nodes Whether a submarine cable is laid between them, The cable type code indicates the number of... In the topology corresponding to each individual, the nodes and The type and specifications of the submarine cable selected.

[0213] Mutation operation:

[0214]

[0215] In the formula: For the first Individuals at any moment The mutated individual vector is the new individual generated by this mutation operation. It will then compete with the original individual through crossover and selection operations to determine whether it enters the next generation of the population. It involves randomly selecting three distinct individual vectors from the population. The core of traditional mutation is... As basis vectors, It is a difference vector that enables random perturbation and evolution of individuals, ensuring the global exploration capability of the population; This is the differential evolution scaling factor, used to control the perturbation step size of the differential vector; The number of wind turbines in the clustering partition is adjusted by reinforcement learning based on the current clustering balance and partitioning cost. It is used to dynamically adjust the clustering labels of individuals and optimize the partitioning structure. =0.1 is the cluster number adjustment coefficient; =0.05 is the hydrogen energy utilization ratio adjustment coefficient.

[0216] Cross operation:

[0217]

[0218] In the formula: For the first Individual, the first The decision variable at time t is the experimental individual vector, which is a new individual generated by the crossover operation. It then competes with the original individual through the selection operation to determine whether to enter the next generation of the population. The dimension is completely consistent with the individual vector and the mutated individual. For the first Individual, the first The mutated individual vectors of the dimensional variables carry optimization information such as the mutated topology, clustering, and hydrogen energy; For the first Individual, the first The original individual vector of the dimensional variable is the th individual in the current population. Preserve the original genes of each individual, retain high-quality genetic information, and avoid population dispersal; This represents the differential evolution crossover probability. The larger the probability, the higher the probability of retaining mutated genes and the stronger the population diversity. Optimize the core decision-making dimensions for the integration of power collection systems and hydrogen energy.

[0219] After completing the architecture design of the reinforcement learning and differential evolution modules, it is necessary to define the distribution range of the wind-hydrogen dual uncertainty through mathematical constraints, providing a unified robust optimization boundary for the aforementioned algorithm modules. At the same time, the results of the previous dual uncertainty modeling are transformed into executable model constraints. Therefore, the following differential robust constraints are set:

[0220]

[0221] In the formula: The set of feasible regions with multi-norm constraints defines the set of all power flow solutions that satisfy the constraints during the optimization process. It is the feasible search space of the optimization algorithm, ensuring that all iterative solutions are within the engineering allowable range. The total system power, which includes all system power variables such as wind farm output power, power flow of the collection system, and load power, is the core decision-making object for optimization. This represents the initial probability distribution for the wind power scenario. This represents the initial probability distribution for hydrogen energy demand scenarios. =0.1、 =0.05 is the probability deviation limit for wind power scenarios; =0.15 is the probability deviation limit for hydrogen energy demand scenarios.

[0222] This constraint is not only the mathematical implementation of the double uncertainty set modeling mentioned above, but also the basis for defining the feasible region of probability distribution in the main and sub-problems of the subsequent column and constraint generation algorithm. It can ensure that the optimization process and subsequent solution steps of the RLDEA algorithm are executed within a unified uncertainty distribution range, thus ensuring the robustness and logical consistency of the entire optimization model.

[0223] To efficiently solve complex mixed-integer linear programming models while ensuring the solution results conform to the fluctuation requirements of dual-uncertainty modeling, the process of using a column and constraint generation algorithm is carried out throughout, combining the dual-uncertainty set obtained from the previous modeling and typical scenarios. Specifically, it includes:

[0224] Main issues (investment in power collection systems + decision-making regarding hydrogen energy access points):

[0225]

[0226] in These are variables for investment decisions, including submarine cable topology, submarine cable type, location and capacity of hydrogen energy access point, and substation site selection; The feasible domain for investment decision variables includes constraints such as submarine cable routing, node capacity, investment budget, and geographical environment, which limits the engineering feasibility of the investment plan. These are the decision variables for operation, including wind turbine output scheduling, power flow distribution of the power collection system, power scheduling of the electrolyzer, and control of wind curtailment, and are the decision objects for lower-level optimization. For a given investment decision The feasible region of the decision variables under the following conditions includes power balance constraints, submarine cable current carrying capacity constraints, node voltage constraints, electrolyzer operating constraints, multi-norm power constraints, etc., and is determined by the investment plan. Decide; For the construction cost of the power collection system; For the operation and maintenance costs of the power collection system; To maximize robustness in the worst-case scenario, for all wind power output probability distributions that satisfy the multi-norm constraint, the worst-case scenario (maximum cost / minimum benefit) is selected to ensure that the optimization scheme is still feasible under the most unfavorable conditions, thereby improving the robustness of the system. To minimize the optimal operating decision, given an investment plan Under the worst probability distribution, take the optimal operating decision. Minimize net operating costs and maximize system operating economy; As a probability expectation operator, it takes the expectation of the probability distribution of wind power output uncertainty, and quantifies the expected operating costs and benefits under uncertainty; The system operation loss cost includes wind curtailment loss cost, grid loss energy cost, and electricity purchase supplement cost, which is the main cost item during the operation phase. The unit value coefficient for hydrogen energy; The efficiency of hydrogen production in the electrolyzer; This refers to the amount of wind power used for on-site hydrogen production, which is the amount of wind power directly supplied to the electrolyzer, corresponding to the amount of wind power consumed on-site.

[0227] Sub-problem (operational optimization, including wind-hydrogen power allocation):

[0228]

[0229] In the formula: A collection of wind power scenarios; A collection of hydrogen energy demand scenarios; The joint probability of wind-hydrogen coupling scenarios; This is a vector of operating cost coefficients in coupled scenarios; For the scene Time period The operational decision variables, corresponding to the discretization of the operational decision y mentioned above, include wind turbine output, electrolyzer power, power flow distribution, and wind curtailment power, which are specific scheduling schemes for each scenario and time period.

[0230] Iteration steps: Initialization =0, feasible region of the main problem Upper Realm The lower realm Solving the main problem yields... Substituting the subproblems into parallel solutions yields the optimal joint probability. and operational efficiency Update upper bound

[0231]

[0232] Generate clipping constraints:

[0233]

[0234] Update feasible domain:

[0235]

[0236] Solve the updated main problem to find the new lower bound. ,like Output the optimal solution; otherwise, k = k + 1 and return to step 2.

[0237] An optimization device for a wind-hydrogen coordinated offshore wind power collection system based on the RLDEA algorithm is disclosed. To implement the aforementioned method, and particularly to ensure the linkage and adaptation between the dual-uncertainty modeling and the RLDEA algorithm, this device employs an optimization method for a wind-hydrogen coordinated offshore wind power collection system based on the RLDEA algorithm of this invention. Specifically, the functional settings of each unit are aligned with the linkage requirements of modeling and the algorithm.

[0238] The target constraint unit is used to establish the core constraints of the power collection system and the collaborative constraints of on-site hydrogen energy utilization, with the core objective of minimizing the total life cycle cost of the power collection system.

[0239] The modeling unit is used to perform dual uncertainty set modeling of wind power output and hydrogen energy demand, and obtain typical wind-hydrogen coupling scenarios through historical data clustering.

[0240] The planning model building unit is used to construct a planning model for a distributed bar collector system based on the RLDEA algorithm, and to achieve simultaneous optimization of clustering partitioning, topology connection, submarine cable selection and hydrogen energy access point.

[0241] Linearization unit, used to transform the model into a mixed-integer linear programming model using the McCormick linearization method;

[0242] The solution unit is used to solve the mixed integer linear programming model using a column and constraint generation algorithm to obtain the optimal power collection system planning scheme.

[0243] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements an optimization method for a wind-hydrogen synergistic offshore wind power collection system based on the RLDEA algorithm of the present invention.

[0244] A storage medium storing a computer program that, when executed by a processor, implements an optimization method for a wind-hydrogen coordinated offshore wind power collection system based on the RLDEA algorithm.

[0245] Example 1:

[0246] This embodiment designs a 35kV voltage-level wind power-hydrogen energy on-site utilization collaborative power collection system for an offshore wind farm with a total installed capacity of 600MW. It deploys 75 8MW wind turbines and 10 electrolyzer access points. Based on the RLDEA algorithm, it completes four-dimensional synchronous optimization of power collection system clustering and partitioning, radial topology design, submarine cable selection, and hydrogen energy access point layout, with the goal of minimizing the discounted cost over the entire life cycle. At the same time, it meets multiple constraints such as safe system operation, collaborative hydrogen energy consumption, and adaptation to wind-hydrogen dual uncertainties. The specific implementation details are as follows.

[0247] Figure 7 This is a diagram of the distributed access architecture for on-site hydrogen energy utilization adopted in this embodiment. After the wind power is collected by the power collection system, part of it is directly connected to the grid and sent out, while the other part is connected to the distributed electrolyzer for hydrogen production nearby, realizing the on-site consumption of abandoned wind power and efficient hydrogen production.

[0248] The basic parameters for this embodiment are set as follows: wind turbine inlet velocity 3 m / s, rated wind speed 12 m / s, outlet wind speed 25 m / s, power factor 0.95~0.96, substation coordinates (6.35×10³m, 2.23×10³m), and wind turbines distributed at... [6.3×10³~6.4×10³m]、 [2.22×10³~2.24×10³m]; Each of the 10 electrolyzer access points is equipped with a 5MW alkaline electrolyzer, with a minimum operating power of 1MW. The access cable is YJV-0.6 / 1kV-3×300, with a unit length resistance of 0.06Ω / km, reactance of 0.08Ω / km, and an average length of 0.8km; the power collection system uses four types of 35kV submarine cables, with a rated current carrying capacity of 250A~512A, a unit length resistance of 0.078~0.342Ω / km, and a cost of 11,190~32,310 yuan / km; the economic parameters are a grid-connected electricity price of 0.5 yuan / kWh, an annual electricity price growth rate of 2%, a discount rate of 8%, a system design life of 25 years, and hydrogen energy. The value coefficient is 1.2 yuan / kWh, the hydrogen production energy conversion efficiency is 0.85, and the unit wind curtailment penalty cost is 45 million yuan / MW·h. The constraint parameters are: system rated voltage 35kV, maximum allowable voltage drop of 5%, maximum voltage drop of 3% for the branch line connected to the electrolyzer, submarine cable transmission power factor of 0.95, minimum on-site utilization ratio of hydrogen energy of 15%, maximum number of feeders of 12, and maximum total power of a single cluster region of 120MW. The RLDEA algorithm population size is 50, the maximum number of iterations is 300, the reinforcement learning learning rate is 0.1, the discount factor is 0.9, the initial value of differential evolution mutation probability is 0.6, the initial value of crossover probability is 0.6, the number of cluster regions is 5, and the column and constraint generation algorithm convergence threshold is 0.01.

[0249] This embodiment takes minimizing the discounted net cost of the entire life cycle of the power collection system as the core objective function, and incorporates the benefits of local hydrogen energy utilization. The complete expression of the objective function is as follows:

[0250]

[0251] Initial construction cost The cost of submarine cable laying, substation construction, and electrolytic cell access equipment is calculated using the following formula:

[0252]

[0253] The calculated cost of the booster station in this embodiment is 85 million yuan, the cost of a single electrolytic cell connection equipment is 1.2 million yuan, the cost per unit length of the connection cable is 8,000 yuan / km, the total laying length of the submarine cable is 236km, the laying length of various types of submarine cables is 40-86km, and the submarine cable connection status variables are... Both are 1, with a total initial construction cost of 189.7 million yuan. The annual power loss cost, taking into account hydrogen diversion and annual electricity price fluctuations, is calculated using the following formula:

[0254]

[0255] Total transmission power of collector branches Based on the wind turbine output and clustering partitioning results, the hydrogen energy diversion power is allocated. Assuming the power loss accounts for 15% to 20% of the total power transmission capacity of the collector branch, the calculated annual power loss cost for the first year is 3.023 million yuan, and the total discounted cost over 25 years is 32.685 million yuan. The annual operation and maintenance cost covers the maintenance of submarine cables, substations, and electrolytic cell access equipment, calculated using the following formula:

[0256]

[0257] The annual operation and maintenance cost of the booster station is 5 million yuan, and the annual operation and maintenance cost of a single electrolyzer connection device is 85,800 yuan, with a stable annual operation and maintenance cost of 8.58 million yuan. The total discounted value over 25 years is 92.156 million yuan. The formula for calculating the annual benefits of on-site hydrogen energy utilization is as follows:

[0258]

[0259] In the first year, the total on-site hydrogen energy utilization is 158,760 MWh, with an annual benefit of 16 million yuan. The total discounted value over 25 years is 172.864 million yuan. Substituting the costs and benefits of each item into the objective function, the discounted net cost of the power collection system over its entire life cycle in this embodiment is calculated to be 189.70 + 32.685 + 92.156 - 172.864 = 141.677 million yuan.

[0260] This embodiment conducts constraint verification from four dimensions: core of the power collection system, synergistic local utilization of hydrogen energy, wind-hydrogen dual uncertainties, and equipment operation linkage. The topological radial constraint is satisfied:

[0261]

[0262] All 75 wind turbines are uniquely connected to the power collection branch, forming 5 radial clusters with the substation as the root node. Each cluster has 14-16 wind turbines, with a total power of 112-128MW per cluster and 10 feeders, all meeting the constraint limits. The submarine cable current carrying capacity constraint is satisfied.

[0263]

[0264] The maximum current transmitted by each collector branch is 382A, with a current carrying capacity utilization rate of 96%. The maximum hydrogen energy shunt current is 152A, which is 38% of the rated current carrying capacity, thus meeting the shunt constraint. The voltage drop constraint is satisfied.

[0265]

[0266] The maximum voltage drop in the main power grid is 4.2%, and the maximum voltage drop in the branch line connecting the electrolytic cell is 2.8%, with no voltage exceeding the limit. The operating power constraints of the electrolytic cell are met.

[0267]

[0268] The operating state variables are consistent with the current collector feeder, and the local hydrogen energy utilization ratio constraint is satisfied:

[0269]

[0270] The utilization rates of the five clustered zones ranged from 16.2% to 20.5%, with an overall utilization rate of 20.5%. Wind power output adopted a piecewise function model, fluctuating between 0 and 8 MW throughout the year, with no output exceeding limits. Hydrogen energy demand was met.

[0271]

[0272] The deviation between forecast and actual demand is controlled within ±20%, and the supply-demand balance deviation rate does not exceed 5%. Equipment operation linkage constraints meet the following requirements:

[0273]

[0274] All current collector branches supplying power to the electrolytic cell remain connected, with a linkage response time of less than 0.5 seconds, meeting the requirements for rapid response.

[0275] This embodiment employs the RLDEA algorithm for synchronous optimization of four-dimensional decision variables in a power collection system. The iterative process is divided into four stages: initialization, adaptive adjustment of reinforcement learning parameters, differential evolution mutation and crossover selection, and convergence judgment. In the initialization stage, 50 decision individuals are randomly generated with a 98-bit encoding length, covering all decision information for 75 wind turbines, 10 electrolyzer access points, and 236km of submarine cable. The reinforcement learning module dynamically adjusts the mutation probability F and crossover probability CR of the differential evolution module based on the state information of each iteration. When the cost reduction rate is high, the constraint satisfaction rate is 100%, and the hydrogen energy utilization rate reaches the target value, the reward function value is positive, and the algorithm maintains the current parameters; otherwise, the parameters are adjusted to increase the search range. After iteration, the mutation probability F stabilizes at 0.75, the crossover probability CR stabilizes at 0.65, and the number of clustered regions remains at 5. The differential evolution module optimizes decision individuals through mutation, crossover, and selection operations. The mutation operation uses:

[0276]

[0277] The crossover operation uses a random crossover method with a crossover probability of 0.65, and the selection operation adopts a greedy strategy. The algorithm reaches the convergence threshold after 287 iterations, which is 31.7% faster than the traditional differential evolution algorithm and 43.7% faster than the genetic algorithm. The optimization results show that the collector system is clustered into 5 optimal partitions, and the topology is a radial ringless network structure.

[0278] The results can be used as a reference. Figure 2 , Figure 2The diagram shows the radial topology of the wind-hydrogen co-current collection system constructed in Example 1. The wind turbines are clustered in zones according to the optimized clustering results, and each branch adopts radial wiring. The hydrogen energy access point and the current collection branch are reasonably matched to balance electrical safety and economy.

[0279] Submarine cable selection follows the principle of "smaller specifications near the end and larger specifications far the end." Submarine cables 1 and 2 are used for near-end wind turbine connections, while cables 3 and 4 are used for far-end wind turbine connections. Hydrogen energy access points are distributed across the power output sections of each collector branch to achieve uniform power distribution. After optimization, the discounted net cost of the collector system over its entire lifecycle is reduced by 13.8% compared to the traditional genetic algorithm, construction costs are reduced by 9.2%, power loss costs are reduced by 15.0%, and operation and maintenance costs are reduced by 7.3%. The solution time is shortened to 890 seconds, a 69.1% reduction compared to the traditional algorithm.

[0280] Figure 4 The figure shows the verification of the constraints on the current carrying capacity of submarine cables and the voltage drop at nodes in Example 1. As can be seen from the figure, the current carrying capacity utilization rate of submarine cables in each current collecting branch and the voltage deviation of the entire network are within the specified limits, which verifies the electrical feasibility of the topology scheme.

[0281] Figure 5 The figure shows a comparison of the iterative convergence of the RLDEA algorithm in Example 1 and the traditional optimization algorithm. It can be seen that the RLDEA algorithm proposed in this invention has a faster convergence speed and higher optimization accuracy, and can achieve a lower total life cycle cost in fewer iterations.

[0282] Example 2:

[0283] This embodiment is designed for a large-scale offshore wind farm with a total installed capacity of 800MW. A wind power-hydrogen energy local utilization collaborative power collection system with a voltage level of 66kV is designed, and 100 8MW wind turbines and 15 electrolyzer access points are deployed.

[0284] The basic parameters for this embodiment are set as follows: the fan parameters are the same as in Embodiment 1, the power factor is 0.96, the substation coordinates are (10×10³m, 8×10³m), and the fans are distributed in... [8×10³~12×10³m]、 [6×10³~10×10³m]; 15 electrolyzer access points are each equipped with a 6MW alkaline electrolyzer, with a minimum operating power of 1.2MW. The access cable is YJV-0.6 / 1kV-3×400, with a unit length resistance of 0.04Ω / km, reactance of 0.07Ω / km, and an average length of 0.9km; the power collection system uses five types of 66kV submarine cables, with a rated current carrying capacity of 400A~720A, a unit length resistance of 0.065~0.180Ω / km, and a cost of 35,200~78,200 RMB / km; the economic parameters are a grid-connected electricity price of 0.52 RMB / kWh, an annual electricity price growth rate of 2%, a discount rate of 7.5%, a system design life of 25 years, and a hydrogen energy value system. The cost per kWh is 1.3 yuan, the hydrogen production energy conversion efficiency is 0.86, and the unit wind curtailment penalty cost is 48 million yuan / MW·h. The constraints are: system rated voltage 66kV, maximum allowable voltage drop of 5%, maximum voltage drop of 3% for the electrolyzer access branch line, submarine cable transmission power factor of 0.96, minimum on-site hydrogen utilization rate of 18%, maximum number of feeders of 15, and maximum total power of a single cluster region of 150MW. The RLDEA algorithm population size is 60, maximum number of iterations is 350, reinforcement learning learning rate is 0.12, discount factor is 0.92, initial value of differential evolution mutation probability is 0.65, initial value of crossover probability is 0.65, number of cluster regions is 6, and column and constraint generation algorithm convergence threshold is 0.008.

[0285] The core objective function of this embodiment is to minimize the discounted net cost of the power collection system over its entire life cycle. Incorporating the benefits of on-site hydrogen energy utilization and considering the characteristics of a 66kV voltage level and a large-scale wind farm, the complete expression of the objective function is:

[0286]

[0287] In this embodiment, the substation construction cost is 150 million yuan, the cost of a single electrolytic cell connection and supporting equipment is 1.5 million yuan, the unit cost of the connection cable is 10,000 yuan / km, the total laying length of the submarine cable is 428km, the laying length of various types of submarine cables is 65~118km, and the submarine cable connection status variables are... All values ​​are 1, and the total initial construction cost is calculated to be 198 million yuan. The formula for calculating the annual power loss cost is:

[0288]

[0289] The benchmark annual electricity price is 0.52 yuan / kWh, and the total transmission power of the power collection branches is... Based on the wind turbine output and the allocation of six clustering partitions, the hydrogen energy diversion power... The power loss is 18%~22% of the power transmission capacity of the collector branch, and even lower at the 66kV voltage level. Calculations show an annual power loss cost of 6.2 million yuan in the first year, and a total discounted cost of 67.582 million yuan over 25 years. The annual operation and maintenance cost calculation formula is as follows:

[0290]

[0291] The annual operation and maintenance cost of the booster station is 12 million yuan, and the annual operation and maintenance cost of a single electrolyzer connection device is 580,000 yuan. The calculated annual operation and maintenance cost is 58 million yuan, with a total discounted value of 628.954 million yuan over 25 years. The formula for calculating the annual benefits of on-site hydrogen energy utilization is:

[0292]

[0293] Calculations show that the total on-site hydrogen energy utilization in the first year is 542,800 MWh, with an annual hydrogen energy utilization benefit of 61.24 million yuan, and a total discounted value of 665.896 million yuan over 25 years. Substituting the costs and benefits of each item into the objective function, the discounted net cost of the power collection system over its entire life cycle in this embodiment is calculated to be 198,000 + 67.582 + 62,895.4 - 66,589.6 = 201,064 million yuan.

[0294] This embodiment also performs multi-dimensional constraint verification, and the topological radial constraint is satisfied:

[0295]

[0296] All 100 wind turbines are uniquely connected to the power collection branch, forming 6 radial clusters with the substation as the root node. Each cluster has 16-17 wind turbines, with a total power output of 128-136MW, all meeting the requirements. =150MW limit requirement, 13 feeders, meeting the maximum limit requirement, feeder load rate uniform. Submarine cable current carrying capacity constraint meets:

[0297]

[0298] The maximum current transmitted by each collector branch is 528A, with a current carrying capacity utilization rate of 82.5%. The maximum hydrogen energy shunt current is 235A, which is 36.7% of the rated current carrying capacity. This meets the shunt constraint and provides greater redundancy, resulting in higher operational safety. The voltage drop constraint is satisfied.

[0299]

[0300] The maximum voltage drop in the main power grid is 4.6%, and the maximum voltage drop in the branch line connecting the electrolytic cell is 2.7%, both far below the limit requirements, indicating excellent voltage quality and eliminating the need for additional reactive power compensation equipment. The electrolytic cell operating power constraints are met.

[0301]

[0302] The operation of the 15 electrolyzers is completely synchronized with their corresponding power collection feeders, with timely response, and the constraint of on-site hydrogen energy utilization ratio is met.

[0303]

[0304] The utilization rates of the six clustered zones ranged from 19.2% to 22.8%, with an overall utilization rate of 22.8%, further enhancing the local absorption capacity of wind power. The wind power output adopted a piecewise function model; the total output after superimposing the outputs of 100 wind turbines fluctuated smoothly within the range of 0 to 800 MW, without significant sudden increases or decreases. This demonstrates stronger resistance to wind power fluctuations, and the hydrogen energy demand is met.

[0305]

[0306] The total power demand from the 15 electrolyzers fluctuates between 0 and 90 MW. The power collection system achieves precise matching between wind power output and hydrogen demand through reasonable power allocation, with a supply-demand balance deviation rate not exceeding 3%. Equipment operation linkage constraints are met.

[0307]

[0308] All current collection branches supplying power to the electrolytic cell are reliably connected. The equipment linkage is remotely controlled by PLC, with a linkage response time of less than 0.3s, meeting the requirements for rapid response and safe operation of the 66kV high-voltage current collection system. The equipment linkage failure rate is less than 0.3%, and the operational stability is significantly improved.

[0309] The RLDEA algorithm iteration process in this embodiment is consistent with that in Embodiment 1. During the initialization phase, 60 decision individuals are generated, each with a 135-bit code, covering all decision information for 100 wind turbines, 15 electrolytic cell access points, and 428km of submarine cable. The reinforcement learning module dynamically adjusts the mutation probability of the differential evolution module based on the optimization requirements of the 66kV power collection system. and crossover probability After multiple generations of iteration, the mutation probability It eventually stabilized at 0.8, with a crossover probability. The algorithm eventually stabilized at 0.7, with the number of clustered regions remaining at 6, achieving optimal adaptation of the algorithm parameters to the 66kV large-scale power collection system. The differential evolution module continuously optimizes decision-making individuals through mutation, crossover, and selection operations. The mutation operation employs:

[0310]

[0311] The crossover operation uses a random crossover method with a crossover probability of 0.7, and the selection operation uses a greedy strategy. The algorithm reaches the convergence threshold after 320 iterations. =0.008, which improves the convergence speed by 33.3% compared to the traditional differential evolution algorithm and by 42.9% compared to the genetic algorithm, significantly improving the solution efficiency. The algorithm optimization results show that the 66kV collector system is clustered into 6 optimal partitions, and the topology is a radial ringless high-voltage collector structure.

[0312] The results can be used as a reference. Figure 3 , Figure 3 The radial topology diagram of the wind-hydrogen co-current collection system in Example 2 shows an optimized partition structure for scenarios with larger installed capacity and higher voltage levels. The current collection network structure is clear, and the hydrogen energy access layout is more concentrated, which is suitable for the needs of large-scale wind-hydrogen co-current operation.

[0313] The submarine cable selection follows the principle of "near-end No. 1, 2, and 3 submarine cables, and far-end No. 4 and 5 submarine cables," fully leveraging the advantages of large-diameter submarine cables in terms of low loss and high current carrying capacity. Hydrogen energy access points are evenly distributed across six clustered partitions, with 2-3 access points in each partition, achieving distributed hydrogen energy utilization. After optimization, the discounted net cost of the power collection system over its entire lifecycle is reduced by 19.8% compared to the traditional genetic algorithm, construction costs are reduced by 9.2%, power loss costs by 27.1%, and operation and maintenance costs by 10.8%. The solution time is shortened to 1240 seconds, a 72.0% reduction compared to the traditional algorithm. Figure 4 This also serves as a verification diagram for the current carrying capacity of the submarine cable and the node voltage drop constraints of the current collection system in Example 2. As can be seen from the diagram, the electrical feasibility of the topology scheme is fully guaranteed.

[0314] Figure 6 The figure shows a comparison of the iterative convergence of the RLDEA algorithm and the traditional algorithm in Example 2. The algorithm still maintains excellent convergence performance in large-scale scenarios, further proving its applicability and superiority.

[0315] Figure 8 The diagram shows the dynamic matching of power between the power collection branch and the electrolyzer over 24 hours. When the wind power output fluctuates, the electrolyzer power can be adjusted in real time to achieve local consumption of wind power and stable hydrogen production, thus verifying the rationality and effectiveness of the wind-hydrogen coordinated operation of this system.

[0316] Please see Figure 9 The diagram shows a structural schematic of a computer device provided in an embodiment of this application. An embodiment of this application provides a computer device 400, including a processor 410 and a memory 420. The memory 420 stores a computer program executable by the processor 410. When the computer program is executed by the processor 410, it performs the method described above.

[0317] This application embodiment also provides a storage medium 430, on which a computer program is stored, and the computer program is executed by a processor 410 to perform the above method.

[0318] The storage medium 430 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0319] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. "A plurality of" means two or more, unless otherwise explicitly specified.

[0320] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0321] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0322] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0323] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0324] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0325] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0326] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

[0327] Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention and not intended to limit it. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the specific embodiments of the present invention, but such modifications or alterations are all within the scope of protection of the pending claims.

Claims

1. An optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA, characterized in that, The process includes the following steps: S1: Establish a core objective function and a multi-dimensional constraint system, construct a full life-cycle discounted cost objective function that includes the time value of money and integrate the benefits of local utilization of hydrogen energy, and simultaneously build a four-dimensional constraint system that integrates electricity, hydrogen energy, dual uncertainties, and equipment linkage to anchor the optimization direction and operating boundary; S2: Wind-hydrogen dual uncertainty modeling, using K-means clustering algorithm to reduce the dimensionality of massive original wind-hydrogen scenarios, generating typical coupled scenarios and joint probabilities, solving the problem of dimensionality explosion in model solution, and significantly reducing the computational load; S3: Construct a planning model for a distributed bluish bar collector system based on the RLDEA algorithm, integrate the advantages of reinforcement learning and differential evolution algorithm, achieve synchronous optimization of four-dimensional decision-making, and improve the scheme's resistance to uncertainty interference by combining distributed bluish bar theory; S4: Model linearization processing: For nonlinear parts such as power loss quadratic terms and uncertainty norm constraints, the model is transformed by introducing auxiliary variables to form a directly computable mixed integer linear programming model. S5: The column and constraint generation algorithm solves the linearized model by splitting it into the main investment decision problem and the sub-problem of operating cost. It generates pruning constraints through iterative interaction, and outputs the optimal planning scheme after satisfying the convergence criterion, forming a complete technical closed loop.

2. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 1, characterized in that, Step S1 focuses on minimizing the life-cycle cost of the power collection system while maximizing the energy utilization rate of hydrogen energy on-site, including: The objective function is to achieve the optimal balance between the total life-cycle cost of the power collection system and the benefits of on-site hydrogen utilization, expressed as: Among them: construction cost of the power collection system : In the formula: The cost per unit length of type b submarine cable; For nodes and Spacing; For nodes and Status of submarine cable connections. =1 indicates a connection, and 0 indicates the opposite; Costs for the construction and expansion of booster stations; This refers to the set of electrolytic cell connection points. For the first The cost of supporting equipment for each electrolytic cell access point; The set of all potential connection edges in a collector system is represented as: in It is a collection of all nodes in the power collection system, including three core types of nodes: wind turbine nodes, offshore substation nodes, and electrolyzer access nodes; Represents a node With nodes An undirected potential connection edge between two nodes represents the engineering possibility of laying a submarine cable between the two nodes; gather It covers all node pairs that meet engineering feasibility requirements, such as no waterway / ecological red line barriers and water depth adaptability, and is the pool of all candidate connection schemes for submarine cable topology optimization; Power loss cost of collector system : In the formula: =8760 represents the number of operating hours per year; For grid connection electricity price; The resistance per unit length of the b-th type of submarine cable; For collector branch Total power; For collector branch Power supplied to the electrolytic cell on-site; Rated voltage; =0.95 is the power factor; Operation and maintenance costs of power collection systems : In the formula: The annual maintenance cost per unit length of the b-type submarine cable; Annual operation and maintenance costs for the booster station; For the first Annual operation and maintenance cost per unit capacity of electrolytic cell; For the first The rated power of each electrolytic cell; Benefits of on-site utilization of hydrogen energy: =1.2 is the hydrogen energy value coefficient; =0.85 represents the energy conversion efficiency of hydrogen production in an electrolyzer; The total annual electricity generated from on-site hydrogen utilization is expressed as: In the formula: For time t, the collector branch Power supplied to the electrolytic cell on-site; =1h is the time step.

3. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 2, characterized in that, To ensure the engineering feasibility and robustness of the planning scheme, and to adapt to the fluctuation characteristics of the initial dual-uncertainty modeling, the established multi-dimensional optimization constraints specifically include the following: all constraints match the boundary of the uncertainty set obtained from the modeling, ensuring that the constraint requirements can cover the fluctuation range of the wind-hydrogen dual sources. Core constraints of the power collection system: topological radiality constraint, submarine cable current carrying capacity constraint, voltage drop constraint, cable crossing avoidance constraint, and clustering partitioning constraint; Coordination constraints for local hydrogen energy utilization: power constraints for electrolyzer access, proportion constraints for local hydrogen energy utilization, power balance constraints between power collection branch and electrolyzer, and linkage constraints between electrolyzer start-up / shutdown and power collection system operation. Dual uncertainty adaptation constraints: power balance constraints and robustness constraints in wind-hydrogen coupling scenarios; Equipment operating status constraints: start-stop linkage constraints between the fan-collector branch-electrolyte cell and the matching constraints between the submarine cable and the electrolyte cell access point.

4. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 3, characterized in that, As the foundation for the safe operation of the system, and to adapt to the fluctuation requirements of dual uncertainty modeling, the core constraints of the power collection system specifically include the following: the limit values ​​of each constraint are set in conjunction with the wind-hydrogen power fluctuation range obtained from the modeling, ensuring that the system can still operate safely even under extreme fluctuation scenarios: Topological radial constraints: In the formula: This refers to the collection of all nodes in the power collection system, that is, all electrical nodes that need to be connected, and includes the following components: This refers to the set of wind turbine nodes, i.e., all wind turbines waiting to be connected to the grid; This refers to the set of nodes for the substation, which is the root node and grid connection hub of the power collection system. This is a set of nodes connected to the electrolyzer, corresponding to the load connection points for on-site hydrogen production from wind power. For booster stations and wind turbines The direct connection state; This represents the maximum number of feeders; Submarine cable current carrying capacity constraints: In the formula: The rated current carrying capacity of type b submarine cable; For collector branch The maximum allowable current allocated to the electrolytic cell satisfies ≤0.4 ; Voltage drop constraint: In the formula: It is a node arrive The total voltage drop of the submarine cable includes the voltage loss caused by the combined effect of wind turbine power generation and hydrogen production load power; For the first Reactance per unit length of submarine cable; For collector branch reactive power; It is the voltage drop component under the sole action of hydrogen production load, characterizing the impact of electrolyzer connection on submarine cable voltage drop, and is used to quantify the disturbance of hydrogen production load on the voltage of current collection system. =5% is the maximum allowable voltage drop of the collector system; This is the distance from the electrolytic cell inlet point to the collector branch node; The resistance per unit length of the cable connecting to the electrolytic cell; =0.69kV is the rated voltage of the electrolytic cell; =3% is the maximum allowable voltage drop when the electrolytic cell is connected; Cable crossing avoidance constraints: In the formula: , This represents the 0-1 connectivity decision variables between two potential submarine cable edges. This indicates that the submarine cable has actually been laid. This indicates that no paving will be carried out; , This represents two intersecting submarine cable edges, signifying a potential connection between two node pairs. These are pre-determined as intersecting edge pairs through planar geometric cross product operations. The set of submarine cables that may intersect is a subset of the set of all potential edges E of the current collector system. It includes all submarine cable edge pairs that will intersect according to geometric determination and is the scope of this constraint. Clustering partitioning constraints: In the formula: For nodes Belongs to the Identifiers of each cluster region; This represents the number of clustered regions.

5. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 3, characterized in that, To achieve efficient on-site hydrogen energy utilization and simultaneously address the wind-hydrogen power fluctuation characteristics modeled with dual uncertainties, the collaborative constraints for on-site hydrogen energy utilization specifically include constraints set in conjunction with typical wind-hydrogen coupling scenarios obtained from modeling, ensuring that hydrogen energy utilization can adapt to the random fluctuations in wind power output: Electrolytic cell access power constraints: In the formula: For the first Total power of the supporting equipment at each electrolytic cell access point The minimum / maximum operating power of the electrolytic cell; For a moment No. The operating status of each electrolytic cell; For a moment No. Total input power of each electrolytic cell; To give the first A collection of current collector branches supplying power to each electrolytic cell; Constraints on the proportion of hydrogen energy used locally: In the formula: For a moment No. The total output of the fans in the clustered region corresponding to each electrolytic cell; =0.15 is the minimum on-site utilization rate of hydrogen energy; Current collector branch - electrolytic cell power balance constraint: In the formula: For a moment collector branch The input power of the fan; For a moment collector branch Power loss; Electrolytic cell start-up / shutdown and current collection system operation linkage constraints: In the formula: For a moment No. The operating status of the current collector feeder corresponding to each electrolytic cell.

6. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 1, characterized in that, The core of the planning model for the sub-blob rod collector system based on the RLDEA algorithm is to establish a linkage mechanism between the modeling results and the algorithm solution. Specifically, all modules of the algorithm are adapted to the output of the previous double uncertainty modeling. Reinforcement learning module: State space: in The construction cost corresponds to the submarine cable construction cost and the equipment cost of the booster station / hydrogen production station in the objective function, representing the economic investment level of the current topology. The better the state, the lower the cost. The network loss cost corresponds to the submarine cable resistance loss and equipment operating loss in the objective function, and characterizes the system operating efficiency. The lower the value, the more energy-efficient and efficient the system is. For submarine cable current, the first The actual current on the submarine cable represents the cable's load rate; it is used to prevent overload and is also the basis for determining whether to switch submarine cable models. For voltage drop, for the first Voltage deviation of the submarine cable; It is a cluster label that identifies the cluster region to which each node belongs, representing the topological partitioning structure; it tells reinforcement learning which partition to optimize in and guides the local search strategy. This represents the discount factor / penalty coefficient, an adaptive parameter within the reinforcement learning algorithm. Let t be the total power of hydrogen energy utilized on-site; Action space: in To enhance the learning module at all times The action space is the decision action vector made by the reinforcement learning agent based on the current state, and it is the core output of the RLDEA algorithm. This is the adjustment amount for the differential evolution scaling factor, used to dynamically control the step size of mutation operations and balance global exploration and local exploitation capabilities; This is the adjustment amount for the crossover probability in differential evolution, used to dynamically adjust the crossover operation probability and optimize the generation efficiency of the topology scheme; This is an adjustment amount for the number of fans in the clustering partition, used to dynamically optimize the clustering balance and adapt to the topology optimization needs of different partitions; This is the adjustment amount for the hydrogen energy weighting coefficient, used for multi-objective optimization that dynamically balances the minimum life-cycle cost of the power collection system with the maximization of the benefits of local hydrogen energy utilization. Reward function: In the formula: =0.6、 =0.2、 =0.1、 =0.1 is the weighting coefficient; The total output of the fan at time t; The number of feasible solutions; The total number of solutions; The total lifecycle cost of the power collection system at time t includes the cost of submarine cable construction, network loss, and operation and maintenance, excluding the benefits of on-site utilization of hydrogen energy. Differential Evolution Module: Individual code: in Optimize the coding for the electrolytic cell connection point. For topology connection encoding, representing the first In the topology corresponding to each individual, the nodes With nodes Whether a submarine cable is laid between them, The cable type code indicates the number of... In the topology corresponding to each individual, the nodes and The type and specifications of the submarine cable selected; Mutation operation: In the formula: For the first Individuals at any moment The mutated individual vector is the new individual generated by this mutation operation. It will then compete with the original individual through crossover and selection operations to determine whether it enters the next generation of the population. It involves randomly selecting three distinct individual vectors from the population. The core of traditional mutation is... As basis vectors, It is a difference vector that enables random perturbation and evolution of individuals, ensuring the global exploration capability of the population; This is the differential evolution scaling factor, used to control the perturbation step size of the differential vector; Adjustment amount for the number of fans in the clustering partition; =0.1 is the cluster number adjustment coefficient; =0.05 is the hydrogen energy utilization ratio adjustment coefficient; Cross operation: In the formula: For the first Individual, the first The decision variable at time t is the experimental individual vector, which is a new individual generated by the crossover operation. It then competes with the original individual through the selection operation to determine whether to enter the next generation of the population. The dimension is completely consistent with the individual vector and the mutated individual. For the first Individual, the first A vector of mutated individuals of a dimensional variable; For the first Individual, the first The original individual vector of the dimensional variable; This represents the differential evolution crossover probability. Optimize core decision-making dimensions for the integration of power collection systems and hydrogen energy; After completing the architecture design of the reinforcement learning and differential evolution modules, it is necessary to define the distribution range of the wind-hydrogen dual uncertainty through mathematical constraints, providing a unified robust optimization boundary for the aforementioned algorithm modules. At the same time, the results of the previous dual uncertainty modeling are transformed into executable model constraints. Therefore, the following differential robust constraints are set: In the formula: The set of feasible regions with multi-norm constraints defines the set of all power / power flow solutions that satisfy the constraints during the optimization process. It is the feasible search space of the optimization algorithm, ensuring that all iterative solutions are within the engineering allowable range. This represents the total power of the system. This represents the initial probability distribution for the wind power scenario. This represents the initial probability distribution for hydrogen energy demand scenarios. =0.1、 =0.05 is the probability deviation limit for wind power scenarios; =0.15 is the probability deviation limit for hydrogen energy demand scenarios.

7. The optimization method for a wind power-hydrogen energy synergistic power collection system based on RLDEA according to claim 1, characterized in that, To efficiently solve complex mixed-integer linear programming models while ensuring the solution results conform to the fluctuation requirements of dual-uncertainty modeling, a column and constraint generation algorithm is used for the solution process. The entire process is conducted in conjunction with the dual-uncertainty set obtained from previous modeling and typical scenarios, specifically including: Main question: in For investment decision variables; The feasible region for investment decision variables; For running decision variables; For a given investment decision The feasible region of the decision variables under running conditions; For the construction cost of the power collection system; For the operation and maintenance costs of the power collection system; To maximize robustness in the worst-case scenario, the worst-case scenario is taken for all wind power output probability distributions that satisfy the multi-norm constraint. To minimize the optimal operating decision, given an investment plan Under the worst probability distribution, take the optimal operating decision. Minimize net operating costs and maximize system operating economy; It is a probability expectation operator; This incurs system operation losses and costs. The unit value coefficient for hydrogen energy; The efficiency of hydrogen production in the electrolyzer; To generate hydrogen on-site from wind power; Sub-problems: In the formula: A collection of wind power scenarios; A collection of hydrogen energy demand scenarios; The joint probability of wind-hydrogen coupling scenarios; This is a vector of operating cost coefficients in coupled scenarios; For the scene Time period The operational decision variables correspond to the discretization of the operational decision y mentioned earlier; Iteration steps: Initialization =0, feasible region of the main problem Upper Realm The lower realm Solving the main problem yields... Substituting the subproblems into parallel solutions yields the optimal joint probability. and operational efficiency Update upper bound Generate clipping constraints: Update feasible domain: Solve the updated main problem to find the new lower bound. ,like Output the optimal solution; otherwise, k = k + 1 and return to S2.

8. A superior device for a wind power-hydrogen energy synergistic collection system based on RLDEA, characterized in that, The device uses the method as described in any one of claims 1 to 7, specifically including that the functional settings of each unit are tailored to the linkage requirements of modeling and algorithms: The target constraint unit is used to establish the core constraints of the power collection system and the collaborative constraints of on-site hydrogen energy utilization, with the core objective of minimizing the total life cycle cost of the power collection system. The modeling unit is used to perform dual uncertainty set modeling of wind power output and hydrogen energy demand, and obtain typical wind-hydrogen coupling scenarios through historical data clustering. The planning model building unit is used to construct a planning model for a distributed bar collector system based on the RLDEA algorithm, and to achieve simultaneous optimization of clustering partitioning, topology connection, submarine cable selection and hydrogen energy access point. Linearization unit, used to transform the model into a mixed-integer linear programming model using the McCormick linearization method; The solution unit is used to solve the mixed integer linear programming model using a column and constraint generation algorithm to obtain the optimal power collection system planning scheme.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-7.

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

Images